AICcmodavg PDF

Package ‘AICcmodavg’
May 29, 2019

Type Package
Title Model Selection and Multimodel Inference Based on (Q)AIC(c)
Version 2.2-2
Date 2019-05-29
Author Marc J. Mazerolle <marc.mazerolle@sbf.ulaval.ca> and portions of code con-
tributed by Dan Linden.
Maintainer Marc J. Mazerolle <marc.mazerolle@sbf.ulaval.ca>
Depends R (>= 3.2.0)
Imports methods, stats, graphics, lattice, MASS, Matrix, nlme, stats4,
survival, unmarked, VGAM, xtable
Suggests betareg, coxme, fitdistrplus, glmmTMB, lavaan, lme4, maxlike,
nnet, ordinal, pscl, R2jags, R2OpenBUGS, R2WinBUGS, jagsUI,
lmerTest
Description
Functions to implement model selection and multimodel inference based on Akaike's informa-
tion criterion (AIC) and the second-order AIC (AICc), as well as their quasi-likelihood counter-
parts (QAIC, QAICc) from various model object classes. The package implements clas-
sic model averaging for a given parameter of interest or predicted values, as well as a shrink-
age version of model averaging parameter estimates or effect sizes. The package includes diag-
nostics and goodness-of-fit statistics for certain model types including those of 'unmarked-
Fit' classes estimating demographic parameters after accounting for imperfect detection probabil-
ities. Some functions also allow the creation of model selection tables for Bayesian mod-
els of the 'bugs', 'rjags', and 'jagsUI' classes. Functions also implement model selection us-
ing BIC. Objects following model selection and multimodel inference can be formatted to La-
TeX using 'xtable' methods included in the package.
License GPL (>= 2)
LazyLoad yes
Repository CRAN
NeedsCompilation no
Date/Publication 2019-05-29 21:20:03 UTC
1
2 R topics documented:
R topics documented:
AICcmodavg-package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
AICc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
AICcCustom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
AICcmodavg-defunct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
aictab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
aictabCustom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
anovaOD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
beetle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
bictab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
bictabCustom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
boot.wt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
bullfrog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
calcium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
cement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
checkConv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
checkParms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
confset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
countDist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
countHist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
covDiag . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
c_hat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
detHist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
DIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
dictab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
dry.frog . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
extractCN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
extractLL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
extractSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
extractX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
fam.link.mer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
fat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
gpa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
ictab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
importance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
iron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
lizards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
mb.gof.test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
min.trap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
modavg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
modavg.utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
modavgCustom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
modavgEffect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
modavgIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
modavgPred . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
modavgShrink . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
AICcmodavg-package 3
multComp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
newt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Nmix.gof.test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
pine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
predictSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
salamander . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
summaryOD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
tortoise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
turkey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
useBIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
useBICCustom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
xtable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
Index 210
AICcmodavg-package Model Selection and Multimodel Inference Based on (Q)AIC(c)
Description
Description: This package includes functions to create model selection tables based on Akaike’s
information criterion (AIC) and the second-order AIC (AICc), as well as their quasi-likelihood
counterparts (QAIC, QAICc). The package also features functions to conduct classic model av-
eraging (multimodel inference) for a given parameter of interest or predicted values, as well as a
shrinkage version of model averaging parameter estimates. Other handy functions enable the com-
putation of relative variable importance, evidence ratios, and confidence sets for the best model.
The present version supports Cox proportional hazards models and conditional logistic regression
(coxph and coxme classes), linear models (lm class), generalized linear models (glm, glm.nb, vglm,
hurdle, and zeroinfl classes), linear models fit by generalized least squares (gls class), linear
mixed models (lme class), generalized linear mixed models (mer, merMod, and glmmTMB classes),
multinomial and ordinal logistic regressions (multinom, polr, clm, and clmm classes), robust re-
gression models (rlm class), beta regression models (betareg class), parametric survival mod-
els (survreg class), nonlinear models (nls and gnls classes), nonlinear mixed models (nlme and
nlmerMod classes), univariate models (fitdist and fitdistr classes), and certain types of latent
variable models (lavaan class). The package also supports various models of unmarkedFit and
maxLikeFit classes estimating demographic parameters after accounting for imperfect detection
probabilities. Some functions also allow the creation of model selection tables for Bayesian models
of the bugs and rjags classes. Objects following model selection and multimodel inference can be
formatted to LaTeX using xtable methods included in the package.
Details
Package: AICcmodavg
Type: Package
Version: 2.2-2
Date: 2019-05-29
License: GPL (>=2 )
LazyLoad: yes
4 AICcmodavg-package
Many functions of the package require a list of models as the input to conduct model selection and
multimodel inference. Thus, you should start by organizing the output of the models in a list (See
’Examples’ below).
This package contains several useful functions for model selection and multimodel inference for
several model classes:
• AICc Computes AIC, AICc, and their quasi-likelihood counterparts (QAIC, QAICc).
• aictab Constructs model selection tables with number of parameters, AIC, delta AIC, Akaike
weights or variants based on AICc, QAIC, and QAICc for a set of candidate models.
• bictab Constructs model selection tables with number of parameters, BIC, delta BIC, BIC
weights for a set of candidate models.
• boot.wt Computes summary statistics from detection histories.
• confset Determines the confidence set for the best model based on one of three criteria.
• DIC Extracts DIC.
• dictab Constructs model selection tables with number of parameters, DIC, delta DIC, DIC
weights for a set of candidate models.
• evidence Computes the evidence ratio between the highest-ranked model based on the infor-
mation criteria selected and a lower-ranked model.
• importance Computes importance values (w+) for the support of a given parameter among
set of candidate models.
• modavg Computes model-averaged estimate, unconditional standard error, and unconditional
confidence interval of a parameter of interest among a set of candidate models.
• modavgEffect Computes model-averaged effect sizes between groups based on the entire
candidate model set.
• modavgShrink Computes shrinkage version of model-averaged estimate, unconditional stan-
dard error, and unconditional confidence interval of a parameter of interest among entire set
of candidate models.
• modavgPred Computes model-average predictions, unconditional SE’s, and confidence inter-
vals among entire set of candidate models.
• multComp Performs multiple comparisons across levels of a factor in a model selection frame-
work.
• useBIC Computes BIC or a quasi-likelihood counterparts (QBIC).
For models not yet supported by the functions above, the following can be useful for model selection
and multimodel inference conducted from input values supplied by the user:
• AICcCustom Computes AIC, AICc, QAIC, and QAICc from user-supplied input values of
log-likelihood and number of parameters.
• aictabCustom Creates model selection tables based on (Q)AIC(c) from user-supplied input
values of log-likelihood and number of parameters.
• bictabCustom Creates model selection tables based on (Q)BIC from user-supplied input val-
ues of log-likelihood and number of parameters.
• ictab Creates model selection tables from user-supplied values of an information criterion.
• modavgCustom Computes model-averaged parameter estimate based on (Q)AIC(c) from user-

supplied input values of log-likelihood, number of parameters, parameter estimates, and stan-
dard errors.
• modavgIC Computes model-averaged parameter estimate from user-supplied values of infor-
mation criterion, parameter estimates, and standard errors.
• useBICCustom Computes BIC and QBIC from user-supplied input values of log-likelihoods
and number of parameters.
A number of functions for model diagnostics are available:
• c_hat Estimates variance inflation factor for binomial or Poisson GLM’s based on various
estimators.
• checkConv Checks the convergence information of the algorithm for the model.
• checkParms Checks the occurrence of parameter estimates with high standard errors in a
model.
• countDist Computes summary statistics from distance sampling data.
• countHist Computes summary statistics from count history data.
• covDiag Computes covariance diagnostics for lambda in N-mixture models.
• detHist Computes summary statistics from detection histories.
• extractCN Extracts condition number from models of certain classes.
• mb.gof.test Computes the MacKenzie and Bailey goodness-of-fit test for single season and
dynamic occupancy models using the Pearson chi-square statistic.
• Nmix.gof.test Computes goodness-of-fit test for N-mixture models based on the Pearson
chi-square statistic.
Other utility functions include:
• anovaOD Computes likelihood-ratio test statistic corrected for overdispersion between two
models.
• extractLL Extracts log-likelihood from models of certain classes.
• extractSE Extracts standard errors from models of certain classes and adds the labels.
• extractX Extracts the predictors and associated information on variables from a list of can-
didate models.
• fam.link.mer Extracts the distribution family and link function from a generalized linear
mixed model of classes mer and merMod.
• predictSE Computes predictions and associated standard errors models of certain classes.
• summaryOD Displays summary of model output adjusted for overdispersion.
• xtable Formats various objects resulting from model selection and multimodel inference to
LaTeX or HTML tables.
Author(s)
Marc J. Mazerolle <marc.mazerolle@uqat.ca>.
References
Anderson, D. R. (2008) Model-based inference in the life sciences: a primer on evidence. Springer:
New York.
Burnham, K. P., and Anderson, D. R. (2002) Model selection and multimodel inference: a practical
information-theoretic approach. Second edition. Springer: New York.
Burnham, K. P., Anderson, D. R. (2004) Multimodel inference: understanding AIC and BIC in
model selection. Sociological Methods and Research 33, 261–304.
Mazerolle, M. J. (2006) Improving data analysis in herpetology: using Akaike’s Information Crite-
rion (AIC) to assess the strength of biological hypotheses. Amphibia-Reptilia 27, 169–180.
Examples
##Example 1: Poisson GLM with offset
##anuran larvae example from Mazerolle (2006)
data(min.trap)
##assign "UPLAND" as the reference level as in Mazerolle (2006)
min.trap$Type <- relevel(min.trap$Type, ref = "UPLAND")
##set up candidate models in a list

Cand.mod <- list()
##global model
Cand.mod[[1]] <- glm(Num_anura ~ Type + log.Perimeter + Num_ranatra,
family = poisson, offset = log(Effort),
data = min.trap)
Cand.mod[[2]] <- glm(Num_anura ~ Type + log.Perimeter, family = poisson,
offset = log(Effort), data = min.trap)
Cand.mod[[3]] <- glm(Num_anura ~ Type + Num_ranatra, family = poisson,
Cand.mod[[4]] <- glm(Num_anura ~ Type, family = poisson,
Cand.mod[[5]] <- glm(Num_anura ~ log.Perimeter + Num_ranatra,
data = min.trap)
Cand.mod[[6]] <- glm(Num_anura ~ log.Perimeter, family = poisson,
Cand.mod[[7]] <- glm(Num_anura ~ Num_ranatra, family = poisson,
Cand.mod[[8]] <- glm(Num_anura ~ 1, family = poisson,
##check c-hat for global model

c_hat(Cand.mod[[1]], method = "pearson") #uses Pearson's chi-square/df
##note the very low overdispersion: in this case, the analysis could be
##conducted without correcting for c-hat as its value is reasonably close
##to 1
##output of model corrected for overdispersion

summaryOD(Cand.mod[[1]], c.hat = 1.04)
##assign names to each model

Modnames <- c("type + logperim + invertpred", "type + logperim",

"type + invertpred", "type", "logperim + invertpred",
"logperim", "invertpred", "intercept only")
##model selection table based on AICc

aictab(cand.set = Cand.mod, modnames = Modnames)
##compute evidence ratio

evidence(aictab(cand.set = Cand.mod, modnames = Modnames))
##compute confidence set based on 'raw' method

confset(cand.set = Cand.mod, modnames = Modnames, second.ord = TRUE,
method = "raw")
##compute importance value for "TypeBOG" - same number of models

##with vs without variable
importance(cand.set = Cand.mod, modnames = Modnames, parm = "TypeBOG")
##compute model-averaged estimate of "TypeBOG" using the natural average

modavg(cand.set = Cand.mod, modnames = Modnames, parm = "TypeBOG")
##compute model-averaged estimate of "TypeBOG" using shrinkage estimator

##same number of models with vs without variable
modavgShrink(cand.set = Cand.mod, modnames = Modnames,
parm = "TypeBOG")
##compute model-averaged predictions for two types of ponds

##create a data set for predictions
dat.pred <- data.frame(Type = factor(c("BOG", "UPLAND")),
log.Perimeter = mean(min.trap$log.Perimeter),
Num_ranatra = mean(min.trap$Num_ranatra),
Effort = mean(min.trap$Effort))
##model-averaged predictions across entire model set

modavgPred(cand.set = Cand.mod, modnames = Modnames,
newdata = dat.pred, type = "response")
##compute model-averaged effect size between two groups

##'newdata' data frame must be limited to two rows
modavgEffect(cand.set = Cand.mod, modnames = Modnames,
newdata = dat.pred, type = "link")
## Not run:
##Example 2: single-season occupancy model example modified from ?occu
require(unmarked)
##single season
data(frogs)
pferUMF <- unmarkedFrameOccu(pfer.bin)
## add some fake covariates for illustration
siteCovs(pferUMF) <- data.frame(sitevar1 = rnorm(numSites(pferUMF)),
sitevar2 = rnorm(numSites(pferUMF)))
## observation covariates are in site-major, observation-minor order

obsCovs(pferUMF) <- data.frame(obsvar1 = rnorm(numSites(pferUMF) *
obsNum(pferUMF)))
##check detection history data from data object

detHist(pferUMF)
##set up candidate model set

fm1 <- occu(~ obsvar1 ~ sitevar1, pferUMF)
##check detection history data from model object
detHist(fm1)
fm2 <- occu(~ 1 ~ sitevar1, pferUMF)

Cand.models <- list(fm1, fm2, fm3, fm4)
##assign names to elements in list

##alternative to using 'modnames' argument
names(Cand.models) <- c("fm1", "fm2", "fm3", "fm4")
##check GOF of global model and estimate c-hat

mb.gof.test(fm4, nsim = 100) #nsim should be > 1000
##check for high SE's in models

lapply(Cand.models, checkParms, simplify = FALSE)
##compute table
print(aictab(cand.set = Cand.models,
second.ord = TRUE), digits = 4)
##export as LaTeX table

if(require(xtable)) {
xtable(aictab(cand.set = Cand.models,
second.ord = TRUE))
}

evidence(aictab(cand.set = Cand.models))
##evidence ratio between top model vs lowest-ranked model
evidence(aictab(cand.set = Cand.models), model.high = "fm2", model.low = "fm3")

confset(cand.set = Cand.models, second.ord = TRUE,
method = "raw")
##compute importance value for "sitevar1" on occupancy

importance(cand.set = Cand.models, parm = "sitevar1",
parm.type = "psi")
##compute model-averaged estimate of "sitevar1" on occupancy

##(natural average)
modavg(cand.set = Cand.models, parm = "sitevar1",

parm.type = "psi")
##compute model-averaged estimate of "sitevar1"

##(shrinkage estimator)
modavgShrink(cand.set = Cand.models,
parm = "sitevar1", parm.type = "psi")
##compute model-average predictions

##check explanatory variables appearing in models
extractX(Cand.models, parm.type = "psi")
##create a data set for predictions

dat.pred <- data.frame(sitevar1 = seq(from = min(siteCovs(pferUMF)$sitevar1),
to = max(siteCovs(pferUMF)$sitevar1), by = 0.5),
sitevar2 = mean(siteCovs(pferUMF)$sitevar2))
##model-averaged predictions of psi across range of values

##of sitevar1 and entire model set
modavgPred(cand.set = Cand.models, newdata = dat.pred,
parm.type = "psi")
detach(package:unmarked)
## End(Not run)
## Not run:
##Example 3: example with user-supplied values of log-likelihoods and
##number of parameters
##vector with model LL's

LL <- c(-38.8876, -35.1783, -64.8970)
##vector with number of parameters

Ks <- c(7, 9, 4)
##create a vector of names to trace back models in set

Modnames <- c("Cm1", "Cm2", "Cm3")
##generate AICc table

aictabCustom(logL = LL, K = Ks, modnames = Modnames, nobs = 121,
sort = TRUE)
##generate AIC table
aictabCustom(logL = LL, K = Ks, modnames = Modnames,
second.ord = FALSE, nobs = 121, sort = TRUE)
##model averaging parameter estimate

##vector of beta estimates for a parameter of interest
model.ests <- c(0.0478, 0.0480, 0.0478)
##vector of SE's of beta estimates for a parameter of interest

model.se.ests <- c(0.0028, 0.0028, 0.0034)
##compute model-averaged estimate and unconditional SE based on AICc

modavgCustom(logL = LL, K = Ks, modnames = Modnames,
estimate = model.ests, se = model.se.ests, nobs = 121)
##compute model-averaged estimate and unconditional SE based on BIC
estimate = model.ests, se = model.se.ests, nobs = 121,
useBIC = TRUE)
## End(Not run)
## Not run:
##Example 4: example with user-supplied values of information criterion
##model selection based on WAIC
##WAIC values
waic <- c(105.74, 107.36, 108.24, 100.57)
##number of effective parameters
effK <- c(7.45, 5.61, 6.14, 6.05)

Modnames <- c("global model", "interactive model",
"additive model", "invertpred model")
##generate WAIC model selection table

ictab(ic = waic, K = effK, modnames = Modnames,
sort = TRUE, ic.name = "WAIC")
##compute model-averaged estimate

##vector of predictions
Preds <- c(0.106, 0.137, 0.067, 0.050)
##vector of SE's for prediction
Ses <- c(0.128, 0.159, 0.054, 0.039)
##compute model-averaged estimate and unconditional SE based on WAIC

modavgIC(ic = waic, K = effK, modnames = Modnames,
estimate = Preds, se = Ses,
ic.name = "WAIC")
##export as LaTeX table

##model-averaged estimate and confidence interval
xtable(modavgIC(ic = waic, K = effK, modnames = Modnames,
ic.name = "WAIC"))
##model selection table with estimate and SE's from each model
xtable(modavgIC(ic = waic, K = effK, modnames = Modnames,
ic.name = "WAIC"), print.table = TRUE)
}
## End(Not run)
AICc 11
AICc Computing AIC, AICc, QAIC, and QAICc
Description
Functions to compute Akaike’s information criterion (AIC), the second-order AIC (AICc), as well
as their quasi-likelihood counterparts (QAIC, QAICc).
Usage
AICc(mod, return.K = FALSE, second.ord = TRUE, nobs = NULL, ...)
## S3 method for class 'aov'

AICc(mod, return.K = FALSE, second.ord = TRUE,
nobs = NULL, ...)
## S3 method for class 'betareg'

nobs = NULL, ...)
## S3 method for class 'clm'

nobs = NULL, ...)
## S3 method for class 'clmm'

nobs = NULL, ...)
## S3 method for class 'coxme'

nobs = NULL, ...)
## S3 method for class 'coxph'

nobs = NULL, ...)
## S3 method for class 'fitdist'

nobs = NULL, ...)
## S3 method for class 'fitdistr'

nobs = NULL, ...)
## S3 method for class 'glm'

nobs = NULL, c.hat = 1, ...)
12 AICc
## S3 method for class 'glmmTMB'

## S3 method for class 'gls'

nobs = NULL, ...)
## S3 method for class 'gnls'

nobs = NULL, ...)
## S3 method for class 'hurdle'

nobs = NULL, ...)
## S3 method for class 'lavaan'

nobs = NULL, ...)
## S3 method for class 'lm'

nobs = NULL, ...)
## S3 method for class 'lme'

nobs = NULL, ...)
## S3 method for class 'lmekin'

nobs = NULL, ...)
## S3 method for class 'maxlikeFit'

## S3 method for class 'mer'

nobs = NULL, ...)
## S3 method for class 'merMod'

nobs = NULL, ...)
## S3 method for class 'lmerModLmerTest'

nobs = NULL, ...)
AICc 13
## S3 method for class 'multinom'

## S3 method for class 'negbin'

nobs = NULL, ...)
## S3 method for class 'nlme'

nobs = NULL, ...)
## S3 method for class 'nls'

nobs = NULL, ...)
## S3 method for class 'polr'

nobs = NULL, ...)
## S3 method for class 'rlm'

nobs = NULL, ...)
## S3 method for class 'survreg'

nobs = NULL, ...)
## S3 method for class 'unmarkedFit'

## S3 method for class 'vglm'

## S3 method for class 'zeroinfl'

nobs = NULL, ...)
Arguments
mod an object of class aov, betareg, clm, clmm, clogit, coxme, coxph, fitdist,
fitdistr, glm, glmmTMB, gls, gnls, hurdle, lavaan, lm, lme, lmekin, maxlikeFit,
mer, merMod, lmerModLmerTest, multinom, negbin, nlme, nls, polr, rlm,
survreg, vglm, zeroinfl, and various unmarkedFit classes containing the out-
put of a model.
14 AICc
return.K logical. If FALSE, the function returns the information criterion specified. If
TRUE, the function returns K (number of estimated parameters) for a given model.
second.ord logical. If TRUE, the function returns the second-order Akaike information crite-
rion (i.e., AICc).
nobs this argument allows to specify a numeric value other than total sample size to
compute the AICc (i.e., nobs defaults to total number of observations). This
is relevant only for mixed models or various models of unmarkedFit classes
where sample size is not straightforward. In such cases, one might use total
number of observations or number of independent clusters (e.g., sites) as the
value of nobs.
c.hat value of overdispersion parameter (i.e., variance inflation factor) such as that ob-
tained from c_hat. Note that values of c.hat different from 1 are only appropri-
ate for binomial GLM’s with trials > 1 (i.e., success/trial or cbind(success, fail-
ure) syntax), with Poisson GLM’s, single-season occupancy models (MacKen-
zie et al. 2002), dynamic occupancy models (MacKenzie et al. 2003), or N-
mixture models (Royle 2004, Dail and Madsen 2011). If c.hat > 1, AICc will
return the quasi-likelihood analogue of the information criteria requested and
multiply the variance-covariance matrix of the estimates by this value (i.e., SE’s
are multiplied by sqrt(c.hat)). This option is not supported for generalized
linear mixed models of the mer or merMod classes.
... additional arguments passed to the function.
Details
AICc computes one of the following four information criteria:
Akaike’s information criterion (AIC, Akaike 1973),
−2 ∗ log − likelihood + 2 ∗ K,
where the log-likelihood is the maximum log-likelihood of the model and K corresponds to the
number of estimated parameters.
Second-order or small sample AIC (AICc, Sugiura 1978, Hurvich and Tsai 1991),
−2 ∗ log − likelihood + 2 ∗ K ∗ (n/(n − K − 1)),
where n is the sample size of the data set.
Quasi-likelihood AIC (QAIC, Burnham and Anderson 2002),
−2 ∗ log − likelihood
QAIC = + 2 ∗ K,
c − hat
where c-hat is the overdispersion parameter specified by the user with the argument c.hat.
Quasi-likelihood AICc (QAICc, Burnham and Anderson 2002),
QAIC = + 2 ∗ K ∗ (n/(n − K − 1))
c − hat
.
Note that AIC and AICc values are meaningful to select among gls or lme models fit by maximum
likelihood. AIC and AICc based on REML are valid to select among different models that only
differ in their random effects (Pinheiro and Bates 2000).
AICc 15
Value
AICc returns the AIC, AICc, QAIC, or QAICc, or the number of estimated parameters, depending
on the values of the arguments.
Note
The actual (Q)AIC(c) values are not really interesting in themselves, as they depend directly on the
data, parameters estimated, and likelihood function. Furthermore, a single value does not tell much
about model fit. Information criteria become relevant when compared to one another for a given
data set and set of candidate models.
Author(s)
Marc J. Mazerolle
References
Akaike, H. (1973) Information theory as an extension of the maximum likelihood principle. In:
Second International Symposium on Information Theory, pp. 267–281. Petrov, B.N., Csaki, F.,
Eds, Akademiai Kiado, Budapest.
Anderson, D. R. (2008) Model-based Inference in the Life Sciences: a primer on evidence. Springer:
New York.
Burnham, K. P., Anderson, D. R. (2002) Model Selection and Multimodel Inference: a practical
Dail, D., Madsen, L. (2011) Models for estimating abundance from repeated counts of an open
population. Biometrics 67, 577–587.
Hurvich, C. M., Tsai, C.-L. (1991) Bias of the corrected AIC criterion for underfitted regression
and time series models. Biometrika 78, 499–509.
MacKenzie, D. I., Nichols, J. D., Lachman, G. B., Droege, S., Royle, J. A., Langtimm, C. A.
(2002) Estimating site occupancy rates when detection probabilities are less than one. Ecology 83,
2248–2255.
MacKenzie, D. I., Nichols, J. D., Hines, J. E., Knutson, M. G., Franklin, A. B. (2003) Estimating
site occupancy, colonization, and local extinction when a species is detected imperfectly. Ecology
84, 2200–2207.
Pinheiro, J. C., Bates, D. M. (2000) Mixed-effect models in S and S-PLUS. Springer Verlag: New
York.
Royle, J. A. (2004) N-mixture models for estimating population size from spatially replicated
counts. Biometrics 60, 108–115.
Sugiura, N. (1978) Further analysis of the data by Akaike’s information criterion and the finite
corrections. Communications in Statistics: Theory and Methods A7, 13–26.
16 AICcCustom
See Also
AICcCustom, aictab, confset, importance, evidence, c_hat, modavg, modavgShrink, modavgPred,

useBIC,
Examples
##cement data from Burnham and Anderson (2002, p. 101)
data(cement)
##run multiple regression - the global model in Table 3.2
glob.mod <- lm(y ~ x1 + x2 + x3 + x4, data = cement)
##compute AICc with full likelihood

AICc(glob.mod, return.K = FALSE)
##compute AIC with full likelihood

AICc(glob.mod, return.K = FALSE, second.ord = FALSE)
##note that Burnham and Anderson (2002) did not use full likelihood
##in Table 3.2 and that the MLE estimate of the variance was
##rounded to 2 digits after decimal point
##compute AICc for mixed model on Orthodont data set in Pinheiro and
##Bates (2000)
## Not run:
require(nlme)
m1 <- lme(distance ~ age, random = ~1 | Subject, data = Orthodont,
method= "ML")
AICc(m1, return.K = FALSE)
## End(Not run)
AICcCustom Compute AIC, AICc, QAIC, and QAICc from User-supplied Input
Description
This function computes Akaike’s information criterion (AIC), the second-order AIC (AICc), as well
as their quasi-likelihood counterparts (QAIC, QAICc) from user-supplied input instead of extract-
ing the values automatically from a model object. This function is particularly useful for output
imported from other software or for model classes that are not currently supported by AICc.
Usage
AICcCustom(logL, K, return.K = FALSE, second.ord = TRUE, nobs = NULL,

c.hat = 1)
AICcCustom 17
Arguments
logL the value of the model log-likelihood.
K the number of estimated parameters in the model.
rion (i.e., AICc).
nobs the sample size required to compute the AICc or QAICc.
ure) syntax), with Poisson GLM’s, single-season or dynamic occupancy models
(MacKenzie et al. 2002, 2003), N-mixture models (Royle 2004, Dail and Mad-
sen 2011), or capture-mark-recapture models (e.g., Lebreton et al. 1992). If
c.hat > 1, AICcCustom will return the quasi-likelihood analogue of the informa-
tion criterion requested.
Details
AICcCustom computes one of the following four information criteria:
Akaike’s information criterion (AIC, Akaike 1973), the second-order or small sample AIC (AICc,
Sugiura 1978, Hurvich and Tsai 1991), the quasi-likelihood AIC (QAIC, Burnham and Anderson
2002), and the quasi-likelihood AICc (QAICc, Burnham and Anderson 2002).
Value
AICcCustom returns the AIC, AICc, QAIC, or QAICc, or the number of estimated parameters,
depending on the values of the arguments.
Note
The actual (Q)AIC(c) values are not really interesting in themselves, as they depend directly on the
Author(s)
Marc J. Mazerolle
References
Akaike, H. (1973) Information theory as an extension of the maximum likelihood principle. In:
Second International Symposium on Information Theory, pp. 267–281. Petrov, B.N., Csaki, F.,
Eds, Akademiai Kiado, Budapest.
18 AICcmodavg-defunct
Hurvich, C. M., Tsai, C.-L. (1991) Bias of the corrected AIC criterion for underfitted regression
and time series models. Biometrika 78, 499–509.
Lebreton, J.-D., Burnham, K. P., Clobert, J., Anderson, D. R. (1992) Modeling survival and testing
biological hypotheses using marked animals: a unified approach with case-studies. Ecological
Monographs 62, 67–118.
2248–2255.
84, 2200–2207.
Sugiura, N. (1978) Further analysis of the data by Akaike’s information criterion and the finite
corrections. Communications in Statistics: Theory and Methods A7, 13–26.
See Also
AICc, aictabCustom, confset, evidence, c_hat, modavgCustom
Examples
data(cement)
##extract log-likelihood
LL <- logLik(glob.mod)[1]
##extract number of parameters

K.mod <- coef(glob.mod) + 1
##compute AICc with full likelihood

AICcCustom(LL, K.mod, nobs = nrow(cement))
AICcmodavg-defunct Defunct Functions in AICcmodavg Package
Description
The functions listed below have been removed from the AICcmodavg package.
AICcmodavg-defunct 19
Usage
AICc.mult(...)
AICc.unmarked(...)
extract.LL(...)
extract.LL.coxph(...)
extract.LL.unmarked(...)
aictab.clm(...)
aictab.clmm(...)
aictab.coxph(...)
aictab.glm(...)
aictab.gls(...)
aictab.lm(...)
aictab.lme(...)
aictab.mer(...)
aictab.merMod(...)
aictab.mult(...)
aictab.nlme(...)
aictab.nls(...)
aictab.polr(...)
aictab.rlm(...)
aictab.unmarked(...)
dictab.bugs(...)
dictab.rjags(...)
modavg.clm(...)
modavg.clmm(...)
modavg.coxph(...)
modavg.glm(...)
modavg.gls(...)
modavg.lme(...)
modavg.mer(...)
modavg.merMod(...)
modavg.mult(...)
modavg.polr(...)
modavg.rlm(...)
modavg.unmarked(...)
modavg.effect(...)
modavg.effect.glm(...)
modavg.effect.gls(...)
modavg.effect.lme(...)
modavg.effect.mer(...)
modavg.effect.merMod(...)
modavg.effect.rlm(...)
modavg.effect.unmarked(...)
modavg.shrink(...)
modavg.shrink.clm(...)
modavg.shrink.clmm(...)
modavg.shrink.coxph(...)
modavg.shrink.glm(...)
modavg.shrink.gls(...)
modavg.shrink.lme(...)
modavg.shrink.mer(...)
modavg.shrink.merMod(...)
modavg.shrink.mult(...)
modavg.shrink.polr(...)
modavg.shrink.rlm(...)
modavg.shrink.unmarked(...)
modavgpred(...)
modavgpred.glm(...)
modavgpred.gls(...)
modavgpred.lme(...)
modavgpred.mer(...)
modavgpred.merMod(...)
modavgpred.rlm(...)
modavgpred.unmarked(...)
mult.comp(...)
predictSE.zip(...)
Arguments
... arguments passed to the function.
Details
AICc.mult has been replaced by AICc.multinom.
AICc.unmarked has been replaced by AICc.unmarkedFit.
extract.LL has been replaced by extractLL.
extract.LL.coxph has been replaced by extractLL.coxph.
extract.LL.unmarked has been replaced by extractLL.unmarkedFit.
aictab.clm has been replaced by aictab.AICsclm.clm.
aictab.clmm has been replaced by aictab.AICclmm.
aictab.coxph has been replaced by aictab.AICcoxph.
aictab.glm has been replaced by aictab.AICglm.lm.
aictab.gls has been replaced by aictab.AICgls.
aictab.lm has been replaced by aictab.AIClm.
aictab.lme has been replaced by aictab.AIClme.
aictab.mer has been replaced by aictab.AICmer.
aictab.merMod has been replaced by aictab.AIClmerMod, aictab.AICglmerMod, or aictab.AICnlmerMod,
depending on the class of the objects.
aictab.mult has been replaced by aictab.AICmultinom.nnet.
aictab.nlme has been replaced by aictab.AICnlme.
aictab.nls has been replaced by aictab.AICnls.
aictab.polr has been replaced by aictab.AICpolr.
AICcmodavg-defunct 21
aictab.rlm has been replaced by aictab.AICrlm.lm.

aictab.unmarked has been replaced by aictab.AICunmarkedFitOccu, aictab.AICunmarkedFitColExt,
aictab.AICunmarkedFitOccuRN, aictab.AICunmarkedFitPCount, aictab.AICunmarkedFitPCO,
aictab.AICunmarkedFitDS, aictab.AICunmarkedFitGDS, aictab.AICunmarkedFitOccuFP, aictab.AICunmarkedFitMP
aictab.AICunmarkedFitGMM, or aictab.AICunmarkedFitGPC, depending on the class of the ob-
jects.
dictab.bugs has been replaced by dictab.AICbugs.
dictab.jags has been replaced by dictab.AICjags.
modavg.clm has been replaced by modavg.AICsclm.clm.
modavg.clmm has been replaced by modavg.AICsclm.clm.
modavg.coxph has been replaced by modavg.AICcoxph.
modavg.glm has been replaced by modavg.AIClm or modavg.AICglm.lm, depending on the class
of the objects.
modavg.gls has been replaced by modavg.AICgls.
modavg.lme has been replaced by modavg.AIClme.
modavg.mer has been replaced by modavg.AICmer.
modavg.merMod has been replaced by modavg.AIClmerMod or modavg.AICglmerMod, depending
on the class of the objects.
modavg.mult has been replaced by modavg.AICmultinom.nnet.
modavg.polr has been replaced by modavg.AICpolr.
modavg.rlm has been replaced by modavg.AICrlm.lm.
modavg.unmarked has been replaced by modavg.AICunmarkedFitOccu, modavg.AICunmarkedFitColExt,
modavg.AICunmarkedFitOccuRN, modavg.AICunmarkedFitPCount, modavg.AICunmarkedFitPCO,
modavg.AICunmarkedFitDS, modavg.AICunmarkedFitGDS, modavg.AICunmarkedFitOccuFP, modavg.AICunmarkedFitMP
modavg.AICunmarkedFitGMM, or modavg.AICunmarkedFitGPC, depending on the class of the ob-
jects.
modavg.effect has been replaced by modavgEffect.
modavg.effect.glm has been replaced by modavgEffect.AICglm.lm or modavgEffect.AIClm,
modavg.effect.gls has been replaced by modavgEffect.AICgls.
modavg.effect.lme has been replaced by modavgEffect.AIClme.
modavg.effect.mer has been replaced by modavgEffect.AICmer.
modavg.effect.merMod has been replaced by modavgEffect.AICglmerMod or modavgEffect.AIClmerMod,
modavg.effect.rlm has been replaced by modavgEffect.AICrlm.lm.
modavg.effect.unmarked has been replaced by modavgEffect.AICunmarkedFitOccu, modavgEffect.AICunmarkedFitCo
modavgEffect.AICunmarkedFitOccuRN, modavgEffect.AICunmarkedFitPCount, modavgEffect.AICunmarkedFitPCO,
modavgEffect.AICunmarkedFitDS, modavgEffect.AICunmarkedFitGDS, modavgEffect.AICunmarkedFitOccuFP,
modavgEffect.AICunmarkedFitMPois, modavgEffect.AICunmarkedFitGMM, or modavgEffect.AICunmarkedFitGPC,
modavg.shrink has been replaced by modavgShrink.
modavg.shrink.clm has been replaced by modavgShrink.AICsclm.clm.

modavg.shrink.clmm has been replaced by modavgShrink.AICclmm.
modavg.shrink.coxph has been replaced by modavgShrink.AICcoxph.
modavg.shrink.glm has been replaced by modavgShrink.AICglm.lm or modavgShrink.AICglm.lm,
modavg.shrink.gls has been replaced by modavgShrink.AICgls.
modavg.shrink.lme has been replaced by modavgShrink.AIClme.
modavg.shrink.mer has been replaced by modavgShrink.AICmer.
modavg.shrink.merMod has been replaced by modavgShrink.AICglmerMod or modavgShrink.AIClmerMod,
modavg.shrink.mult has been replaced by modavgShrink.AICmultinom.nnet.
modavg.shrink.polr has been replaced by modavgShrink.AICpolr.
modavg.shrink.rlm has been replaced by modavgShrink.AICrlm.lm
modavg.shrink.unmarked has been replaced by modavgShrink.AICunmarkedFitOccu, modavgShrink.AICunmarkedFitCo
modavgShrink.AICunmarkedFitOccuRN, modavgShrink.AICunmarkedFitPCount, modavgShrink.AICunmarkedFitPCO,
modavgShrink.AICunmarkedFitDS, modavgShrink.AICunmarkedFitGDS, modavgShrink.AICunmarkedFitOccuFP,
modavgShrink.AICunmarkedFitMPois, modavgShrink.AICunmarkedFitGMM, or modavgShrink.AICunmarkedFitGPC,
modavgpred has been replaced by modavgPred.
modavgpred.glm has been replaced by modavgpred.AICglm.lm or modavgPred.AIClm, depending
on the class of the objects.
modavgpred.gls has been replaced by modavgPred.AICgls.
modavgpred.lme has been replaced by modavgPred.AIClme.
modavgpred.mer has been replaced by modavgPred.AICmer.
modavgpred.merMod has been replaced by modavgpred.AICglmerMod or modavgPred.AIClmerMod,
modavgpred.rlm has been replaced by modavgPred.AICrlm.lm.
modavgpred.unmarked has been replaced by modavgPred.AICunmarkedFitOccu, modavgPred.AICunmarkedFitColExt,
modavgPred.AICunmarkedFitOccuRN, modavgPred.AICunmarkedFitPCount, modavgPred.AICunmarkedFitPCO,
modavgPred.AICunmarkedFitDS, modavgPred.AICunmarkedFitGDS, modavgPred.AICunmarkedFitOccuFP,
modavgPred.AICunmarkedFitMPois, modavgPred.AICunmarkedFitGMM, or modavgPred.AICunmarkedFitGPC,
mult.comp has been replaced by multComp.
predictSE.zip has been replaced by predictSE.
Author(s)
Marc J. Mazerolle
See Also
aictab, confset, dictab, importance, evidence, extractLL, c_hat, modavg, modavgEffect,
modavgShrink, modavgPred, multComp, predictSE
aictab 23
aictab Create Model Selection Tables
Description
This function creates a model selection table based on one of the following information criteria:
AIC, AICc, QAIC, QAICc. The table ranks the models based on the selected information criteria
and also provides delta AIC and Akaike weights. aictab selects the appropriate function to create
the model selection table based on the object class. The current version works with lists containing
objects of aov, betareg, clm, clmm, clogit, coxme, coxph, fitdist, fitdistr, glm, glmmTMB,
gls, gnls, hurdle, lavaan, lm, lme, lmekin, maxlikeFit, mer, merMod, lmerModLmerTest,
multinom, negbin, nlme, nls, polr, rlm, survreg, vglm, and zeroinfl classes as well as var-
ious models of unmarkedFit classes but does not yet allow mixing of different classes.
Usage
aictab(cand.set, modnames = NULL, second.ord = TRUE, nobs = NULL,
sort = TRUE, ...)
## S3 method for class 'AICaov.lm'

aictab(cand.set, modnames = NULL,
second.ord = TRUE, nobs = NULL, sort = TRUE, ...)
## S3 method for class 'AICbetareg'

## S3 method for class 'AICsclm.clm'

## S3 method for class 'AICclmm'

## S3 method for class 'AICclm'

## S3 method for class 'AICcoxme'

## S3 method for class 'AICcoxph'

24 aictab
## S3 method for class 'AICfitdist'

## S3 method for class 'AICfitdistr'

## S3 method for class 'AICglm.lm'

second.ord = TRUE, nobs = NULL, sort = TRUE, c.hat = 1, ...)
## S3 method for class 'AICglmmTMB'

## S3 method for class 'AICgls'

## S3 method for class 'AICgnls.gls'

## S3 method for class 'AIChurdle'

## S3 method for class 'AIClavaan'

## S3 method for class 'AIClm'

## S3 method for class 'AIClme'

## S3 method for class 'AIClmekin'

## S3 method for class 'AICmaxlikeFit.list'

aictab 25
## S3 method for class 'AICmer'

## S3 method for class 'AIClmerMod'

## S3 method for class 'AIClmerModLmerTest'

## S3 method for class 'AICglmerMod'

## S3 method for class 'AICnlmerMod'

## S3 method for class 'AICmultinom.nnet'

## S3 method for class 'AICnegbin.glm.lm'

## S3 method for class 'AICnlme.lme'

## S3 method for class 'AICnls'

## S3 method for class 'AICpolr'

## S3 method for class 'AICrlm.lm'

## S3 method for class 'AICsurvreg'

26 aictab
## S3 method for class 'AICunmarkedFitOccu'

## S3 method for class 'AICunmarkedFitColExt'

## S3 method for class 'AICunmarkedFitOccuRN'

## S3 method for class 'AICunmarkedFitPCount'

## S3 method for class 'AICunmarkedFitPCO'

## S3 method for class 'AICunmarkedFitDS'

## S3 method for class 'AICunmarkedFitGDS'

## S3 method for class 'AICunmarkedFitOccuFP'

## S3 method for class 'AICunmarkedFitMPois'

## S3 method for class 'AICunmarkedFitGMM'

## S3 method for class 'AICunmarkedFitGPC'

## S3 method for class 'AICunmarkedFitOccuMulti'

aictab 27
## S3 method for class 'AICvglm'

## S3 method for class 'AICzeroinfl'

Arguments
cand.set a list storing each of the models in the candidate model set.
modnames a character vector of model names to facilitate the identification of each model
in the model selection table. If NULL, the function uses the names in the cand.set
list of candidate models (i.e., a named list). If no names appear in the list and
no character vector is provided, generic names (e.g., Mod1, Mod2) are supplied
in the table in the same order as in the list of candidate models.
rion (i.e., AICc).
value of nobs.
sort logical. If TRUE, the model selection table is ranked according to the (Q)AIC(c)
values.
mixture models (Royle 2004, Dail and Madsen 2011). If c.hat > 1, aictab
will return the quasi-likelihood analogue of the information criteria requested
and multiply the variance-covariance matrix of the estimates by this value (i.e.,
SE’s are multiplied by sqrt(c.hat)). This option is not supported for general-
ized linear mixed models of the mer or merMod classes.
Details
aictab internally creates a new class for the cand.set list of candidate models, according to the
contents of the list. The current function is implemented for clogit, coxme, coxph, fitdist,
fitdistr, glm, glmmTMB, gls, gnls, hurdle, lavaan, lm, lme, lmekin, maxlikeFit, mer, merMod,
lmerModLmerTest, multinom, negbin, nlme, nls, polr, rlm, survreg, vglm, and zeroinfl classes
as well as various unmarkedFit classes.
The function constructs a model selection table based on one of the four information criteria: AIC,
AICc, QAIC, and QAICc.
28 aictab
Ten guidelines for model selection:

1) Carefully construct your candidate model set. Each model should represent a specific (interest-
ing) hypothesis to test.
2) Keep your candidate model set short. It is ill-advised to consider as many models as there are
data.
3) Check model fit. Use your global model (most complex model) or subglobal models to determine
if the assumptions are valid. If none of your models fit the data well, information criteria will only
indicate the most parsimonious of the poor models.
4) Avoid data dredging (i.e., looking for patterns after an initial round of analysis).
5) Avoid overfitting models. You should not estimate too many parameters for the number of ob-
servations available in the sample.
6) Be careful of missing values. Remember that values that are missing only for certain variables
change the data set and sample size, depending on which variable is included in any given model. I
suggest to remove missing cases before starting model selection.
7) Use the same response variable for all models of the candidate model set. It is inappropriate to
run some models with a transformed response variable and others with the untransformed variable.
A workaround is to use a different link function for some models (e.g., identity vs log link).
8) When dealing with models with overdispersion, use the same value of c-hat for all models in
the candidate model set. For binomial models with trials > 1 (i.e., success/trial or cbind(success,
failure) syntax) or with Poisson GLM’s, you should estimate the c-hat from the most complex
model (global model). If c-hat > 1, you should use the same value for each model of the candidate
model set (where appropriate) and include it in the count of parameters (K). Similarly, for negative
binomial models, you should estimate the dispersion parameter from the global model and use the
same value across all models.
9) Burnham and Anderson (2002) recommend to avoid mixing the information-theoretic approach
and notions of significance (i.e., P values). It is best to provide estimates and a measure of their
precision (standard error, confidence intervals).
10) Determining the ranking of the models is just the first step. Akaike weights sum to 1 for the
entire model set and can be interpreted as the weight of evidence in favor of a given model being
the best one given the candidate model set considered and the data at hand. Models with large
Akaike weights have strong support. Evidence ratios, importance values, and confidence sets for
the best model are all measures that assist in interpretation. In cases where the top ranking model
has an Akaike weight > 0.9, one can base inference on this single most parsimonious model. When
many models rank highly (i.e., delta (Q)AIC(c) < 4), one should model-average effect sizes for the
parameters with most support across the entire set of models. Model averaging consists in making
inference based on the whole set of candidate models, instead of basing conclusions on a single
’best’ model. It is an elegant way of making inference based on the information contained in the
entire model set.
Value
aictab creates an object of class aictab with the following components:
Modname the name of each model of the candidate model set.

K the number of estimated parameters for each model.
aictab 29
(Q)AIC(c) the information criterion requested for each model (AIC, AICc, QAIC, QAICc).
Delta_(Q)AIC(c)
the appropriate delta AIC component depending on the information criteria se-
lected.
ModelLik the relative likelihood of the model given the data (exp(-0.5*delta[i])). This is
not to be confused with the likelihood of the parameters given the data. The
relative likelihood can then be normalized across all models to get the model
probabilities.
(Q)AIC(c)Wt the Akaike weights, also termed "model probabilities" sensu Burnham and An-
derson (2002) and Anderson (2008). These measures indicate the level of sup-
port (i.e., weight of evidence) in favor of any given model being the most parsi-
monious among the candidate model set.
Cum.Wt the cumulative Akaike weights. These are only meaningful if results in table are
sorted in decreasing order of Akaike weights (i.e., sort = TRUE).
c.hat if c.hat was specified as an argument, it is included in the table.
LL if c.hat = 1 and parameters estimated by maximum likelihood, the log-likelihood
of each model.
Quasi.LL if c.hat > 1, the quasi log-likelihood of each model.
Res.LL if parameters are estimated by restricted maximum-likelihood (REML), the re-
stricted log-likelihood of each model.
Author(s)
Marc J. Mazerolle
References
New York.
2248–2255.
84, 2200–2207.
30 aictab
See Also
AICc, aictabCustom, bictab, confset, c_hat, evidence, importance, modavg, modavgEffect,
modavgShrink, modavgPred
Examples
##Mazerolle (2006) frog water loss example
data(dry.frog)
##setup a subset of models of Table 1

Cand.models <- list( )
Cand.models[[1]] <- lm(log_Mass_lost ~ Shade + Substrate +
cent_Initial_mass + Initial_mass2,
data = dry.frog)
cent_Initial_mass + Initial_mass2 +
Shade:Substrate, data = dry.frog)
Cand.models[[3]] <- lm(log_Mass_lost ~ cent_Initial_mass +
Initial_mass2, data = dry.frog)
Cand.models[[4]] <- lm(log_Mass_lost ~ Shade + cent_Initial_mass +
Cand.models[[5]] <- lm(log_Mass_lost ~ Substrate + cent_Initial_mass +

Modnames <- paste("mod", 1:length(Cand.models), sep = " ")

aictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE)
##round to 4 digits after decimal point and give log-likelihood
print(aictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE),
digits = 4, LL = TRUE)
## Not run:
##Burnham and Anderson (2002) flour beetle data
data(beetle)
##models as suggested by Burnham and Anderson p. 198
Cand.set <- list( )
Cand.set[[1]] <- glm(Mortality_rate ~ Dose, family =
binomial(link = "logit"), weights = Number_tested,
data = beetle)
binomial(link = "probit"), weights = Number_tested,
data = beetle)
binomial(link ="cloglog"), weights = Number_tested,
data = beetle)
##check c-hat
c_hat(Cand.set[[1]])
aictab 31
##lowest value of c-hat < 1 for these non-nested models, thus use
##c.hat = 1
##set up named list

names(Cand.set) <- c("logit", "probit", "cloglog")
##compare models
##model names will be taken from the list if modnames is not specified
res.table <- aictab(cand.set = Cand.set, second.ord = FALSE)
##note that delta AIC and Akaike weights are identical to Table 4.7
print(res.table, digits = 2, LL = TRUE) #print table with 2 digits and
##print log-likelihood in table
print(res.table, digits = 4, LL = FALSE) #print table with 4 digits and
##do not print log-likelihood
## End(Not run)
##two-way ANOVA with interaction

data(iron)
##full model
m1 <- lm(Iron ~ Pot + Food + Pot:Food, data = iron)
##additive model
m2 <- lm(Iron ~ Pot + Food, data = iron)
##null model
m3 <- lm(Iron ~ 1, data = iron)
##candidate models
Cand.aov <- list(m1, m2, m3)
Cand.names <- c("full", "additive", "null")
aictab(Cand.aov, Cand.names)
##single-season occupancy model example modified from ?occu

## Not run:
require(unmarked)
##single season example modified from ?occu
data(frogs)
##add fake covariates
sitevar2 = runif(numSites(pferUMF)))
##observation covariates
obsNum(pferUMF)))

32 aictabCustom

##assemble models in named list (alternative to using 'modnames' argument)

Cand.mods <- list("fm1" = fm1, "fm2" = fm2, "fm3" = fm3, "fm4" = fm4)
##compute table
aictab(cand.set = Cand.mods, second.ord = TRUE)
## End(Not run)
aictabCustom Create Model Selection Tables from User-supplied Input Based on

(Q)AIC(c)
Description
This function creates a model selection table from model input (log-likelihood, number of estimated
parameters) supplied by the user instead of extracting the values automatically from a list of can-
didate models. The models are ranked based on one of the following information criteria: AIC,
AICc, QAIC, QAICc. The table ranks the models based on the selected information criteria and
also provides delta AIC and Akaike weights.
Usage
aictabCustom(logL, K, modnames = NULL, second.ord = TRUE, nobs = NULL,
sort = TRUE, c.hat = 1)
Arguments
logL a vector of log-likelihood values for the models in the candidate model set.
K a vector containing the number of estimated parameters for each model in the
rion (i.e., AICc).
values.
aictabCustom 33
c.hat > 1, aictabCustom will return the quasi-likelihood analogue of the infor-
mation criterion requested.
Details
aictabCustom constructs a model selection table based on one of the four information criteria:
AIC, AICc, QAIC, and QAICc. This function is most useful when model input is imported into
R from other software (e.g., Program MARK, PRESENCE) or for model classes that are not yet
supported by aictab.
Value
aictabCustom creates an object of class aictab with the following components:

(Q)AIC(c) the information criteria requested for each model (AICc, AICc, QAIC, QAICc).
Delta_(Q)AIC(c)
lected.
probabilities.
Cum.Wt the cumulative Akaike weights. These are only meaningful if results in table are
sorted in decreasing order of Akaike weights (i.e., sort = TRUE).
of each model.
Author(s)
Marc J. Mazerolle
34 anovaOD
References
New York.
2248–2255.
84, 2200–2207.
See Also
AICcCustom, bictabCustom, confset, c_hat, evidence, ictab, modavgCustom
Examples
LL <- c(-38.8876, -35.1783, -64.8970)

Ks <- c(7, 9, 4)


aictabCustom(logL = LL, K = Ks, modnames = Modnames, nobs = 121,
sort = TRUE)
anovaOD Likelihood-Ratio Test Corrected for Overdispersion
Description
Compute likelihood-ratio test between a given model and a simpler model.
anovaOD 35
Usage
anovaOD(mod.simple, mod.complex, c.hat = 1,

nobs = NULL, ...)

nobs = NULL, ...)
## S3 method for class 'unmarkedFitOccu'

nobs = NULL, ...)
## S3 method for class 'unmarkedFitColExt'

nobs = NULL, ...)
## S3 method for class 'unmarkedFitOccuRN'

nobs = NULL, ...)
## S3 method for class 'unmarkedFitPCount'

nobs = NULL, ...)
## S3 method for class 'unmarkedFitPCO'

nobs = NULL, ...)
## S3 method for class 'unmarkedFitDS'

nobs = NULL, ...)
## S3 method for class 'unmarkedFitGDS'

nobs = NULL, ...)
## S3 method for class 'unmarkedFitOccuFP'

nobs = NULL, ...)
## S3 method for class 'unmarkedFitMPois'

nobs = NULL, ...)
## S3 method for class 'unmarkedFitGMM'

nobs = NULL, ...)
36 anovaOD
## S3 method for class 'unmarkedFitGPC'

nobs = NULL, ...)
## S3 method for class 'glmerMod'

nobs = NULL, ...)

nobs = NULL, ...)

nobs = NULL, ...)

nobs = NULL, ...)
Arguments
mod.simple an object of class glm, glmmTMB, maxlikeFit, mer, merMod, multinom, vglm,
and various unmarkedFit classes containing the output of a model. This model
should be a simpler version of mod.complex resulting from a deletion of certain
terms (i.e., nested model).
mod.complex an object of the same class as mod.simple.
c.hat value of overdispersion parameter (i.e., variance inflation factor) such as that
obtained from c_hat, mb.gof.test, or Nmix.gof.test. Typically, this value
should be computed for the most complex model and applied to simpler models.
nobs the number of observations used in the analysis. If nobs = NULL, the total
number of rows are used as the sample size to compute the residual degrees of
freedom as nobs − K, where K is the number of estimated parameters. This
value of nobs.
Details
This function applies a correction for overdispersion on the likelihood-ratio test between a model
and its simpler counterpart. The simpler model must be nested within the more complex model,
typically as the result of deleting terms. You should supply the c.hat value of the most complex of
the two models you are comparing.
When 1 < c.hat < 4, the likelihood-ratio test is computed as:
anovaOD 37
−2 ∗ (LL.simple − LL.complex)
LR =
(K.complex − K.simple) ∗ c.hat
where LL.simple and LL.complex are the log-likelihoods of the simple and complex models, respec-
tively, and where K.complex and K.simple are the number of estimated parameters in each model.
The test statistic is approximately distributed as FK.complex−K.simple,n−K.complex , where n is the
number of observations (i.e., nobs) used in the analysis (Venables and Ripley 2002).
When nobs = NULL, the number of observations is based on the number of rows of the data frame
used in the analysis. For mixed models or various models of unmarkedFit, sample size is less
straightforward, and nobs could be based on the total number of observations or on the number of
independent clusters (e.g., sites), among other choices.
When c.hat = 1, the likelihood-ratio test simplifies to:
LR = −2 ∗ (LL.simple − LL.complex)
where in this case the test statistic is distributed as a χ2K.complex−K.simple (McCullagh and Nelder
1989).
The function supports different model types such as Poisson GLM’s and GLMM’s, single-season
and dynamic occupancy models (MacKenzie et al. 2002, 2003), and various N-mixture models
(Royle 2004, Dail and Madsen 2011).
Value
anovaOD returns an object of class anovaOD as a list with the following components:
form.simple a character string of the parameters estimated in mod.simple.

form.complex a character string of the parameters estimated in mod.complex.
c.hat the c.hat estimate used to adjust the likelihood-ratio test.
devMat a matrix storing as columns the number of parameters estimated (K), the log-
likelihood of each model logLik, the difference in estimated parameters be-
tween the two models (Kdiff), minus twice the difference in log-likelihoods
between the models (-2LL), the test statistic, and the associated P-value.
Author(s)
Marc J. Mazerolle
References
2248–2255.
38 anovaOD
84, 2200–2207.
McCullagh, P., Nelder, J. A. (1989) Generalized Linear Models. Second edition. Chapman and
Hall: New York.
Venables, W. N., Ripley, B. D. (2002) Modern Applied Statistics with S. Second edition. Springer-
Verlag: New York.
See Also
c_hat, mb.gof.test, Nmix.gof.test, summaryOD
Examples
data(min.trap)
##run model
m1 <- glm(Num_anura ~ Type + log.Perimeter + Num_ranatra,
data = min.trap)
##null model
m0 <- glm(Num_anura ~ 1,
data = min.trap)

c_hat(m1) #uses Pearson's chi-square/df
##likelihood ratio test corrected for overdispersion

anovaOD(mod.simple = m0, mod.complex = m1, c.hat = c_hat(m1))
##compare without overdispersion correction
anovaOD(mod.simple = m0, mod.complex = m1)
##example with occupancy model

## Not run:
##load unmarked package
if(require(unmarked)){
data(bullfrog)
##detection data
detections <- bullfrog[, 3:9]
beetle 39
##assemble in unmarkedFrameOccu
bfrog <- unmarkedFrameOccu(y = detections)
##run model
fm <- occu(~ 1 ~ Reed.presence, data = bfrog)
##null model
fm0 <- occu(~ 1 ~ 1, data = bfrog)
##check GOF
##GOF <- mb.gof.test(fm, nsim = 1000)
##estimate of c-hat: 1.89
##display results after overdispersion adjustment

anovaOD(fm0, fm, c.hat = 1.89)
}
## End(Not run)
beetle Flour Beetle Data
Description
This data set illustrates the acute mortality of flour beetles (Tribolium confusum) following 5 hour
exposure to carbon disulfide gas.
Usage
data(beetle)
Format
A data frame with 8 rows and 4 variables.
Dose dose of carbon disulfide in mg/L.

Number_tested number of beetles exposed to given dose of carbon disulfide.
Number_killed number of beetles dead after 5 hour exposure to given dose of carbon disulfide.
Mortality_rate proportion of total beetles found dead after 5 hour exposure.
Details
Burnham and Anderson (2002, p. 195) use this data set originally from Young and Young (1998) to
show model selection for binomial models with different link functions (logit, probit, cloglog).
40 bictab
Source
Young, L. J., Young, J. H. (1998) Statistical Ecology. Kluwer Academic Publishers: London.
Examples
data(beetle)
## maybe str(beetle) ; plot(beetle) ...
bictab Create Model Selection Tables Based on BIC
Description
This function creates a model selection table based on the Bayesian information criterion (Schwarz
1978, Burnham and Anderson 2002). The table ranks the models based on the BIC and also provides
delta BIC and BIC model weights. The function adjusts for overdispersion in model selection by
using the QBIC when c.hat > 1. bictab selects the appropriate function to create the model
selection table based on the object class. The current version works with lists containing objects of
aov, betareg, clm, clmm, clogit, coxme, coxph, fitdist, fitdistr, glm, glmmTMB, gls, gnls,
hurdle, lavaan, lm, lme, lmekin, maxlikeFit, mer, merMod, lmerModLmerTest, multinom, nlme,
nls, polr, rlm, survreg, vglm, and zeroinfl classes as well as various models of unmarkedFit
classes but does not yet allow mixing of different classes.
Usage
bictab(cand.set, modnames = NULL, nobs = NULL,
sort = TRUE, ...)

bictab(cand.set, modnames = NULL,
nobs = NULL, sort = TRUE, ...)




bictab 41



## S3 method for class 'AICfitdist'

## S3 method for class 'AICfitdistr'


nobs = NULL, sort = TRUE, c.hat = 1, ...)


## S3 method for class 'AICgnls.gls'


## S3 method for class 'AIClavaan'



42 bictab









## S3 method for class 'AICnlme.lme'

## S3 method for class 'AICnls'



bictab 43













44 bictab




Arguments
compute the BIC (i.e., nobs defaults to total number of observations). This is
relevant only for mixed models or various models of unmarkedFit classes where
sample size is not straightforward. In such cases, one might use total number of
observations or number of independent clusters (e.g., sites) as the value of nobs.
sort logical. If TRUE, the model selection table is ranked according to the BIC values.
mixture models (Royle 2004, Dail and Madsen 2011). If c.hat > 1, bictab
will return the quasi-likelihood analogue of the BIC (QBIC) and multiply the
variance-covariance matrix of the estimates by this value (i.e., SE’s are mul-
tiplied by sqrt(c.hat)). This option is not supported for generalized linear
mixed models of the mer or merMod classes.
Details
BIC tends to favor simpler models than AIC whenever n > 8 (Schwarz 1978, Link and Barker
2006, Anderson 2008). BIC assigns uniform prior probabilities across all models (i.e., equal 1/R),
whereas in AIC and AICc, prior probabilities increase with sample size (Burnham and Anderson
2004, Link and Barker 2010). Some authors argue that BIC requires the true model to be included
bictab 45
in the model set, whereas AIC or AICc does not (Burnham and Anderson 2002). However, Link
and Barker (2006, 2010) consider both as assuming that a model in the model set approximates
truth.
bictab internally creates a new class for the cand.set list of candidate models, according to the
contents of the list. The current function is implemented for clogit, coxme, coxph, fitdist,
fitdistr, glm, glmmTMB, gls, gnls, hurdle, lavaan, lm, lme, lmekin, maxlikeFit, mer, merMod,
lmerModLmerTest, multinom, nlme, nls, polr, rlm, survreg, vglm, and zeroinfl classes as well
as various unmarkedFit classes. The function constructs a model selection table based on BIC.
Value
bictab creates an object of class bictab with the following components:

(Q)BIC the Bayesian information criterion for each model.
Delta_(Q)BIC the delta BIC component.
probabilities.
(Q)BICWt the BIC model weights, also termed "model probabilities" (Burnham and An-
derson 2002, Link and Barker 2006, Anderson 2008). These measures indicate
the level of support (i.e., weight of evidence) in favor of any given model being
the most parsimonious among the candidate model set.
Cum.Wt the cumulative BIC weights. These are only meaningful if results in table are
sorted in decreasing order of BIC weights (i.e., sort = TRUE).
LL the log-likelihood of each model.
Res.LL if parameters are estimated by restricted maximum-likelihood (REML), the re-
stricted log-likelihood of each model.
Author(s)
Marc J. Mazerolle
References
New York.
46 bictab
Link, W. A., Barker, R. J. (2006) Model weights and the foundations of multimodel inference.
Ecology 87, 2626–2635.
Link, W. A., Barker, R. J. (2010) Bayesian Inference with Ecological Applications. Academic
Press: Boston.
2248–2255.
84, 2200–2207.
Schwarz, G. (1978) Estimating the dimension of a model. Annals of Statistics 6, 461–464.
See Also
aictab, bictabCustom, confset, evidence, importance, useBIC,
Examples
data(dry.frog)

data = dry.frog)

##generate BIC table

bictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE)
##round to 4 digits after decimal point and give log-likelihood
print(bictab(cand.set = Cand.models, modnames = Modnames, sort = TRUE),
digits = 4, LL = TRUE)
## Not run:
bictab 47

data(beetle)
Cand.set <- list( )
data = beetle)
data = beetle)
data = beetle)
##set up named list

names(Cand.set) <- c("logit", "probit", "cloglog")
##compare models
##model names will be taken from the list if modnames is not specified
bictab(cand.set = Cand.set)
## End(Not run)

data(iron)
##full model
m1 <- lm(Iron ~ Pot + Food + Pot:Food, data = iron)
##additive model
m2 <- lm(Iron ~ Pot + Food, data = iron)
##null model
m3 <- lm(Iron ~ 1, data = iron)
##candidate models
Cand.aov <- list(m1, m2, m3)
Cand.names <- c("full", "additive", "null")
bictab(Cand.aov, Cand.names)

## Not run:
require(unmarked)
##single season example modified from ?occu
data(frogs)
##add fake covariates
##observation covariates
obsNum(pferUMF)))
48 bictabCustom

##assemble models in named list (alternative to using 'modnames' argument)

Cand.mods <- list("fm1" = fm1, "fm2" = fm2, "fm3" = fm3, "fm4" = fm4)
##compute table based on QBIC that accounts for c.hat

bictab(cand.set = Cand.mods, c.hat = 3.9)
## End(Not run)
bictabCustom Create Model Selection Tables from User-supplied Input Based on

(Q)BIC
Description
This function creates a model selection table from model input (log-likelihood, number of estimated
parameters) supplied by the user instead of extracting the values automatically from a list of can-
didate models. The models are ranked based on the BIC (Schwarz 1978) or on a quasi-likelihood
analogue (QBIC) corrected for overdispersion. The table ranks the models based on the selected
information criteria and also provides delta BIC and BIC weights.
Usage
bictabCustom(logL, K, modnames = NULL, nobs = NULL, sort = TRUE,
c.hat = 1)
Arguments
sort logical. If TRUE, the model selection table is ranked according to the (Q)BIC
values.
bictabCustom 49
c.hat > 1, bictabCustom will return the quasi-likelihood analogue of the infor-
Details
bictabCustom constructs a model selection table based on BIC or QBIC. This function is most use-
ful when model input is imported into R from other software (e.g., Program MARK, PRESENCE)
or for model classes that are not yet supported by bictab.
Value
bictabCustom creates an object of class bictab with the following components:

(Q)BIC the information criteria requested for each model (BIC, QBIC).
Delta_(Q)BIC the appropriate delta BIC component depending on the information criteria se-
lected.
probabilities.
(Q)BICWt the BIC weights, also termed "model probabilities" sensu Burnham and Ander-
son (2002) and Anderson (2008). These measures indicate the level of support
(i.e., weight of evidence) in favor of any given model being the most parsimo-
nious among the candidate model set.
Cum.Wt the cumulative BIC weights. These are only meaningful if results in table are
sorted in decreasing order of BIC weights (i.e., sort = TRUE).
of each model.
Author(s)
Marc J. Mazerolle
50 bictabCustom
References
New York.
2248–2255.
84, 2200–2207.
See Also
AICcCustom, aictabCustom, confset, c_hat, evidence, ictab, modavgCustom
Examples

LL <- c(-38.8876, -35.1783, -64.8970)

Ks <- c(7, 9, 4)

##generate BIC table

bictabCustom(logL = LL, K = Ks, modnames = Modnames, nobs = 121,
sort = TRUE)
boot.wt 51
boot.wt Compute Model Selection Relative Frequencies
Description
This function computes the model selection relative frequencies based on the nonparametric boot-
strap (Burnham and Anderson 2002). Models are ranked based on the AIC, AICc, QAIC, or QAICc.
The function currently supports objects of aov, betareg, clm, glm, hurdle, lm, multinom, polr,
rlm, survreg, vglm, and zeroinfl classes.
Usage
boot.wt(cand.set, modnames = NULL, second.ord = TRUE, nobs = NULL,
sort = TRUE, nsim = 100, ...)




sort = TRUE, nsim = 100, c.hat = 1, ...)




52 boot.wt




Arguments
list of candidate models. If no names appear in the list, generic names (e.g.,
Mod1, Mod2) are supplied in the table in the same order as in the list of candidate
models.
rion (i.e., AICc).
is relevant only for certain types of models such as mixed models where sam-
ple size is not straightforward. In such cases, one might use total number of
values.
obtained from c_hat. Note that values of c.hat different from 1 are only appro-
priate for binomial GLM’s with trials > 1 (i.e., success/trial or cbind(success,
failure) syntax) or with Poisson GLM’s. If c.hat > 1, boot.wt will return the
quasi-likelihood analogue of the information criterion requested.
nsim the number of bootstrap iterations. Burnham and Anderson (2002) recommend
at least 1000 and up to 10 000 iterations for certain problems.
Details
boot.wt is implemented for aov, betareg, glm, hurdle, lm, multinom, polr, rlm, survreg, vglm,
and zeroinfl classes. During each bootstrap iteration, the data are resampled with replacement,
all the models specified in cand.set are updated with the new data set, and the top-ranked model is
boot.wt 53
saved. When all iterations are completed, the relative frequency of selection is computed for each
model appearing in the candidate model set.
Relative frequencies of the models are often similar to Akaike weights, and the latter are often
preferred due to their link with a Bayesian perspective (Burnham and Anderson 2002). boot.wt
is most useful for teaching purposes of sampling-theory based relative frequencies of model se-
lection. The current implementation is only appropriate with completely randomized designs. For
more complex data structures (e.g., blocks or random effects), the bootstrap should be modified
accordingly.
Value
boot.wt creates an object of class boot.wt with the following components:
Modname the names of each model of the candidate model set.

(Q)AIC(c) the information criteria requested for each model (AICc, AICc, QAIC, QAICc).
Delta_(Q)AIC(c)
lected.
probabilities.
PiWt the relative frequencies of model selection from the bootstrap.
Author(s)
Marc J. Mazerolle
References
New York.
54 boot.wt
See Also
AICc, confset, c_hat, evidence, importance, modavg, modavgShrink, modavgPred
Examples
data(dry.frog)

data = dry.frog)

##generate AICc table with bootstrapped relative

##frequencies of model selection
boot.wt(cand.set = Cand.models, modnames = Modnames, sort = TRUE,
nsim = 10) #number of iterations should be much higher

## Not run:
data(beetle)
Cand.set <- list( )
data = beetle)
data = beetle)
data = beetle)

Modnames <- paste("Mod", 1:length(Cand.set), sep = " ")
##model selection table with bootstrapped

##relative frequencies
bullfrog 55
boot.wt(cand.set = Cand.set, modnames = Modnames)
## End(Not run)
bullfrog Bullfrog Occupancy and Common Reed Invasion
Description
This is a data set from Mazerolle et al. (2014) on the occupancy of Bullfrogs (Lithobates cates-
beianus) in 50 wetlands sampled in 2009 in the area of Montreal, QC.
Usage
data(bullfrog)
Format
A data frame with 50 observations on the following 23 variables.
Location a factor with a unique identifier for each wetland.
Reed.presence a binary variable, either 1 (reed present) or 0 (reed absent).
V1 a binary variable for detection (1) or non detection (0) of bullfrogs during the first survey.
V2 a binary variable for detection (1) or non detection (0) of bullfrogs during the second survey.
V3 a binary variable for detection (1) or non detection (0) of bullfrogs during the third survey.
V4 a binary variable for detection (1) or non detection (0) of bullfrogs during the fourth survey.
V5 a binary variable for detection (1) or non detection (0) of bullfrogs during the fifth survey.
V6 a binary variable for detection (1) or non detection (0) of bullfrogs during the sixth survey.
V7 a binary variable for detection (1) or non detection (0) of bullfrogs during the seventh survey.
Effort1 a numeric variable for the centered number of sampling stations during the first survey.
Effort2 a numeric variable for the centered number of sampling stations during the second survey.
Effort3 a numeric variable for the centered number of sampling stations during the third survey.
Effort4 a numeric variable for the centered number of sampling stations during the fourth survey.
Effort5 a numeric variable for the centered number of sampling stations during the fifth survey.
Effort6 a numeric variable for the centered number of sampling stations during the sixth survey.
Effort7 a numeric variable for the centered number of sampling stations during the seventh survey.
Type1 a binary variable to identify the survey type, either minnow trap (1) or call survey (0) during
the first sampling occasion.
the second sampling occasion.
the third sampling occasion.
56 calcium
the fourth sampling occasion.
the fifth sampling occasion.
the sixth sampling occasion.
the seventh sampling occasion.
Details
This data set is used to illustrate single-species single-season occupancy models (MacKenzie et al.
2002) in Mazerolle (2015).
Source
(2002). Estimating site occupancy rates when detection probabilities are less than one. Ecology 83,
2248–2255.
Mazerolle, M. J., Perez, A., Brisson, J. (2014) Common reed (Phragmites australis) invasion and
amphibian distribution in freshwater wetlands. Wetlands Ecology and Management 22, 325–340.
Mazerolle, M. J. (2015) Estimating detectability and biological parameters of interest with the use
of the R environment. Journal of Herpetology 49, 541–559.
Examples
data(bullfrog)
str(bullfrog)
calcium Blood Calcium Concentration in Birds
Description
This data set features calcium concentration in the plasma of birds of both sexes following a hor-
monal treatment.
Usage
data(calcium)
Format
Calcium calcium concentration in mg/100 ml in the blood of birds.
Hormone a factor with two levels indicating whether the bird received a hormonal treatment or not.
Sex a factor with two levels coding for the sex of birds.
cement 57
Details
Zar (1984, p. 206) illustrates a two-way ANOVA with interaction with this data set.
Source
Zar, J. H. (1984) Biostatistical analysis. Second edition. Prentice Hall: Englewood Cliffs, New
Jersey.
Examples
data(calcium)
str(calcium)
cement Heat Expended Following Hardening of Portland Cement
Description
This data set illustrates the heat expended (calories) from mixtures of four different ingredients of
Portland cement expressed as a percentage by weight.
Usage
data(cement)
Format
x1 calcium aluminate.
x2 tricalcium silicate.
x3 tetracalcium alumino ferrite.
x4 dicalcium silicate.
y calories of heat per gram of cement following 180 days of hardening.
Details
Burnham and Anderson (2002, p. 101) use this data set originally from Woods et al. (1932) to
select among a set of multiple regression models.
Source
Woods, H., Steinour, H. H., Starke, H. R. (1932) Effect of composition of Portland cement on heat
evolved during hardening. Industrial and Engineering Chemistry 24, 1207–1214.
58 checkConv
Examples
data(cement)
## maybe str(cement) ; plot(cement) ...
checkConv Check Convergence of Fitted Model
Description
This function checks the convergence information contained in models of various classes.
Usage
checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)
checkConv 59

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)

checkConv(mod, ...)
Arguments
mod an object containing the output of a model of the classes mentioned above.
Details
This function checks the element of a model object that contains the convergence information from
the optimization function. The function is currently implemented for models of classes betareg,
clm, clmm, glm, glmmTMB, hurdle, lavaan, maxlikeFit, merMod, lmerModLmerTest, multinom,
nls, polr, unmarkedFit, and zeroinfl. The function is particularly useful for functions with
several groups of parameters, such as those of the unmarked package (Fiske and Chandler, 2011).
Value
checkConv returns a list with the following components:
converged a logical value indicating whether the algorithm converged or not.

message a string containing the message from the optimization function.
Author(s)
Marc J. Mazerolle
References
Fiske, I., Chandler, R. (2011) unmarked: An R Package for fitting hierarchical models of wildlife
occurrence and abundance. Journal of Statistical Software 43, 1–23.
See Also
checkParms, covDiag, mb.gof.test, Nmix.gof.test

60 checkParms
Examples
##example modified from ?pcount
## Not run:
##Simulate data
set.seed(3)
nSites <- 100
nVisits <- 3
##covariate
x <- rnorm(nSites)
beta0 <- 0
beta1 <- 1
##expected counts
lambda <- exp(beta0 + beta1*x)
N <- rpois(nSites, lambda)
y <- matrix(NA, nSites, nVisits)
p <- c(0.3, 0.6, 0.8)
for(j in 1:nVisits) {
y[,j] <- rbinom(nSites, N, p[j])
}
## Organize data
visitMat <- matrix(as.character(1:nVisits),
nSites, nVisits, byrow=TRUE)
umf <- unmarkedFramePCount(y=y, siteCovs=data.frame(x=x),

obsCovs=list(visit=visitMat))
## Fit model
fm1 <- pcount(~ visit ~ 1, umf, K=50)
checkConv(fm1)
}
## End(Not run)
checkParms Identify Parameters with Large Standard Errors
Description
This function identifies parameter estimates with large standard errors in a model. It is particularly
useful for complex models with different parameter types such as those of unmarkedFit classes
implemented in package unmarked (Fiske and Chandler, 2011), as well as other types of regression
models.
Usage
checkParms(mod, se.max = 25, simplify = TRUE, ...)

checkParms(mod, se.max = 25, ...)
checkParms 61
















62 checkParms






checkParms(mod, se.max = 25, simplify = TRUE,
...)


Arguments
mod a model of unmarkedFit classes or other regression model. This model is
checked to determine the occurrence of large standard errors for parameter esti-
mates.
se.max specifies the value beyond which standard errors are deemed high for the model
at hand. The function will determine the number of estimates with standard
errors that exceed se.max.
simplify this argument is only valid for models of unmarkedFit classes which consist of
several parameter types for detection probability and demographic parameters
(e.g., abundance, occupancy, extinction). If TRUE, the function returns a matrix
with a single row identifying the parameter type and estimate with the highest
standard error. If FALSE, the function returns a matrix with as many rows as
there are parameter types in the model. In the latter case, the estimate with the
highest standard error for each parameter type is presented.
Details
In some complex models such as certain hierarchical models (Royle and Dorazio 2008, Kéry and
Royle 2015), issues in estimating parameters and their standard errors can occur. Large standard
checkParms 63
errors can be indicative of problems in estimating certain parameters due to sparse data, parameters
on the boundary, or model misspecification. The checkParms function computes the number of
parameter estimates with standard errors larger than se.max and identifies the parameter estimate
with the largest standard error across all parameter types (simplify = TRUE) or for each parameter
type (simplify = FALSE).
To help identify large standard errors, users can standardize numeric explanatory variables to zero
mean and unit variance. The checkParms function can also be useful to identify boundary estimates
in classic generalized models or their extensions (Venables and Ripley 2002).
Value
checkParms returns a list of class checkParms with the following components:
model.class the class of the model for which diagnostics are requested.
se.max the value of SE used as a threshold in diagnostics. The function reports the
number of parameter estimates with SE > se.max.
result a matrix consisting of three columns, namely, the identity of the parameter es-
timate with the highest SE (variable), its standard error (max.se), and the
number of parameter estimates with SE larger than se.max (n.high.se). For
classical regression models with a single response variable, the row name is la-
beled beta. For unmarkedFit models, the matrix either consists of a single row
(simplify = TRUE) labeled with the name of the parameter type (e.g., psi,
gam, eps, p) where the highest SE occurs, or consists of as many rows as there
are parameter types (simplify = FALSE).
Author(s)
Marc J. Mazerolle
References
Agresti, A. (2002) Categorical data analysis. John Wiley and Sons, Inc.: Hoboken.
Fiske, I., Chandler, R. (2011) unmarked: An R Package for fitting hierarchical models of wildlife
occurrence and abundance. Journal of Statistical Software 43, 1–23.
Kéry, M., Royle, J. A. (2015) Applied hierarchical modeling in ecology: analysis of distribution,
abundance and species richness in R and BUGS. Academic Press, New York, USA.
Royle, J. A., Dorazio, R. M. (2008) Hierarchical modeling and inference in ecology: the analysis
of data from populations, metapopulations and communities. Academic Press: New York.
Venables, W. N., Ripley, B. D. (2002) Modern applied statistics with S, 2nd edition. Springer-
Verlag: New York.
See Also
c_hat, detHist, checkConv, countDist, countHist, extractCN, mb.gof.test, Nmix.gof.test,
parboot
64 confset
Examples
##example with multiple-season occupancy model modified from ?colext
## Not run:
require(unmarked)
data(frogs)
umf <- formatMult(masspcru)
obsCovs(umf) <- scale(obsCovs(umf))
siteCovs(umf) <- rnorm(numSites(umf))
yearlySiteCovs(umf) <- data.frame(year = factor(rep(1:7,
numSites(umf))))
##model with with year-dependent transition rates

fm.yearly <- colext(psiformula = ~ 1, gammaformula = ~ year,
epsilonformula = ~ year,
pformula = ~ JulianDate + I(JulianDate^2),
data = umf)
##check for high SE's and report highest

##across all parameter types
checkParms(fm.yearly, simplify = TRUE)
##check for high SE's and report highest

##for each parameter type
checkParms(fm.yearly, simplify = FALSE)
## End(Not run)
##example from Agresti 2002 of logistic regression

##with parameters estimated at the boundary (complete separation)
## Not run:
x <- c(10, 20, 30, 40, 60, 70, 80, 90)
y <- c(0, 0, 0, 0, 1, 1, 1, 1)
m1 <- glm(y ~ x, family = binomial)

checkParms(m1)
## End(Not run)
confset Computing Confidence Set for the Kullback-Leibler Best Model
Description
This function computes the confidence set on the best model given the data and model set. confset
implements three different methods proposed by Burnham and Anderson (2002).
Usage
confset(cand.set, modnames = NULL, second.ord = TRUE, nobs = NULL,
method = "raw", level = 0.95, delta = 6, c.hat = 1)
confset 65
Arguments
models.
rion (i.e., AICc).
value of nobs.
method a character value, either as raw, ordinal, or ratio, indicating the method for
determining the confidence set for the best model (see ’Description’ above for
details).
level the level of confidence (i.e., sum of model probabilities) used to determine the
confidence set on the best model when using the raw method. Note that the
argument is not used for the other methods of determining the confidence set on
the best model.
delta the delta (Q)AIC(c) value associated with the cutoff point to determine the
confidence set for the best model. Note that the argument is only used when
method = ratio.
mixture models (Royle 2004, Dail and Madsen 2011). If c.hat > 1, confset
Details
The first and simplest (method = 'raw'), relies on summing the Akaike weights (i.e., model
probabilities) of the ranked models until we reach a given cutpoint (e.g., 0.95 for a 95 percent set).
The second method (method = 'ordinal') suggested is based on the classification of the models
on an ordinal scale based on the delta (Q)AIC(c). The models are grouped in different classes based
on their weight of support as determined by the delta (Q)AIC(c) values: substantial support (delta
(Q)AIC(c) <= 2), some support (2 < delta (Q)AIC(c) <= 7), little support (7 < delta (Q)AIC(c) <=
10), no support (delta (Q)AIC(c) > 10).
66 confset
The third method (method = 'ratio') is based on identifying the ratios of model likelihoods
(i.e., exp(-delta_(Q)AIC(c)/2) ) that exceed a cutpoint, similar to the building of profile likelihood
intervals. An evidence ratio of each model relative to the top-ranked model is computed and the
ratios exceeding the cutpoint determine which models are included in the confidence set. Note
here that small cutoff points are suggested (e.g., 0.125, 0.050). The cutoff point is linked to delta
(Q)AIC(c) by the following relationship: cutof f = exp(−1 ∗ δ(Q)AIC(c) /2).
Value
confset returns an object of class confset as a list with the following components, depending on
which method is used:
when method = 'raw':
method identifies the method of determining the confidence set on the best model.
level the confidence level used to determine the confidence set on the best model.
table a reduced table with the models included in the confidence set.
when method = 'ordinal':
substantial a reduced table with the models included in the confidence set for which delta
(Q)AIC(c) <= 2.
some a reduced table with the models included in the confidence set for which 2 <
delta (Q)AIC(c) <= 7.
little a reduced table with the models included in the confidence set for which 7 <
delta (Q)AIC(c) <= 10.
none a reduced table with the models included in the confidence set for which delta
(Q)AIC(c) > 10.
when method = 'ratio':
cutoff the cutoff value for the ratios used to determine the confidence set on the best
model.
delta the delta (Q)AIC(c) used to compute the cutoff value for ratios to determine the
confidence set on the best model.
table a reduced table with the models included in the confidence set.
Author(s)
Marc J. Mazerolle
References
confset 67
2248–2255.
84, 2200–2207.
See Also
AICc, aictab, c_hat, evidence, importance, modavg, modavgShrink, modavgPred
Examples
data(min.trap)
##set up candidate models

Cand.mod <- list()
##global model
Cand.mod[[1]] <- glm(Num_anura ~ Type + log.Perimeter + Num_ranatra,
data = min.trap)
data = min.trap)
Cand.mod[[7]] <- glm(Num_anura ~ Num_ranatra, family = poisson,

c_hat(Cand.mod[[1]]) #uses Pearson's chi-square/df
##to 1

Modnames <- c("type + logperim + invertpred", "type + logperim",
"type + invertpred", "type", "logperim + invertpred",
68 countDist
"logperim", "invertpred", "intercept only")

confset(cand.set = Cand.mod, modnames = Modnames, second.ord = TRUE,
method = "raw")
##example with linear mixed model

## Not run:
require(nlme)
##set up candidate model list for Orthodont data set shown in Pinheiro
##and Bates (2000: Mixed-effect models in S and S-PLUS. Springer Verlag:
##New York.)
Cand.models <- list()
Cand.models[[1]] <- lme(distance ~ age, random = ~age | Subject,
data = Orthodont, method = "ML")
Cand.models[[2]] <- lme(distance ~ age + Sex, data = Orthodont,
random = ~ 1 | Subject, method = "ML")
Cand.models[[3]] <- lme(distance ~ 1, data = Orthodont,
random = ~ 1 | Subject, method = "ML")
##create a vector of model names

Modnames <- paste("mod", 1:length(Cand.models), sep = "")

confset(cand.set = Cand.models, modnames = Modnames, second.ord = TRUE,
method = "raw")
##round to 4 digits after decimal point
print(confset(cand.set = Cand.models, modnames = Modnames,
second.ord = TRUE, method = "raw"), digits = 4)

level = 0.9, method = "raw")
##compute confidence set based on 'ordinal' method

method = "ordinal")
##compute confidence set based on 'ratio' method

method = "ratio", delta = 4)

method = "ratio", delta = 8)
detach(package:nlme)
## End(Not run)
countDist Compute Summary Statistics from Distance Sampling Data

countDist 69
Description
This function extracts various summary statistics from distance sampling data of various unmarkedFrame
and unmarkedFit classes.
Usage
countDist(object, plot.freq = TRUE, plot.distance = TRUE, ...)
## S3 method for class 'unmarkedFrameDS'

countDist(object, plot.freq = TRUE,
plot.distance = TRUE, ...)

## S3 method for class 'unmarkedFrameGDS'


Arguments
object an object of various unmarkedFrame or unmarkedFit classes containing dis-
tance sampling data.
plot.freq logical. Specifies if the count data (pooled across seasons and distance classes)
should be plotted.
plot.distance logical. Specifies if the counts in each distance class should be plotted.
Details
This function computes a number of summary statistics in data sets used for the distance sampling
models of Royle et al. (2004) and Chandler et al. (2011).
countDist can take data frames of the unmarkedFrameDS and unmarkedFrameGDS classes as in-
put. For convenience, the function can also extract the raw data from model objects of classes
unmarkedFitDS and unmarkedFitGDS. Note that different model objects using the same data set
will have identical values.
Value
countDist returns a list with the following components:
70 countDist
count.table.full
a table with the frequency of each observed count pooled across distances classes.
count.table.seasons
a list of tables with the frequency of each season-specific count pooled across
distance classes.
dist.sums.full a table with the frequency of counts in each distance class across the entire
sampling seasons.
hist.table.seasons
a list of tables with the frequency of counts in each distance class for each pri-
mary period.
out.freqs a matrix where the rows correspond to each sampling season and where columns
consist of the number of sites sampled in season t (sampled) and the number of
sites with at least one detection in season t (detected). For multiseason data,
the matrix includes the number of sites sampled in season t − 1 with coloniza-
tions observed in season t (colonized), the number of sites sampled in season
t − 1 with extinctions observed in season t (extinct), the number of sites sam-
pled in season t − 1 without changes observed in season t (static), and the
number of sites sampled in season t that were also sampled in season t − 1
(common).
out.props a matrix where the rows correspond to each sampling season and where columns
consist of the proportion of sites in season t with at least one detection (naive.occ).
For multiseason data, the matrix includes the proportion of sites sampled in sea-
son t − 1 with colonizations observed in season t (naive.colonization), the
proportion of sites sampled in season t − 1 with extinctions observed in season
t (naive.extinction), and the proportion of sites sampled in season t − 1 with
no changes observed in season t.
n.seasons the number of seasons (primary periods) in the data set.
n.visits.season
the maximum number of visits per season in the data set.
Author(s)
Marc J. Mazerolle
References
Chandler, R. B., Royle, J. A., King, D. I. (2011) Inference about density and temporary emigration
in unmarked populations. Ecology 92, 1429–1435.
Royle, J. A., Dawson, D. K., Bates, S. (2004) Modeling abundance effects in distance sampling.
Ecology 85, 1591–1597.
See Also
covDiag, detHist, countHist, Nmix.chisq, Nmix.gof.test

countHist 71
Examples
##modified example from ?distsamp
## Not run:
data(linetran)
##format data
ltUMF <- with(linetran, {
unmarkedFrameDS(y = cbind(dc1, dc2, dc3, dc4),
siteCovs = data.frame(Length, area, habitat),
dist.breaks = c(0, 5, 10, 15, 20),
tlength = linetran$Length * 1000, survey = "line",
unitsIn = "m")
})
##compute descriptive stats from data object

countDist(ltUMF)
##Half-normal detection function

fm1 <- distsamp(~ 1 ~ 1, ltUMF)
##compute descriptive stats from model object
countDist(fm1)
}
## End(Not run)
countHist Compute Summary Statistics from Count Histories
Description
This function extracts various summary statistics from count data of various unmarkedFrame and
unmarkedFit classes.
Usage
countHist(object, plot.freq = TRUE, ...)
## S3 method for class 'unmarkedFramePCount'


## S3 method for class 'unmarkedFrameGPC'


72 countHist
## S3 method for class 'unmarkedFrameMPois'


## S3 method for class 'unmarkedFramePCO'

countHist(object, plot.freq = TRUE,
plot.seasons = FALSE, ...)

## S3 method for class 'unmarkedFrameGMM'


Arguments
object an object of various unmarkedFrame or unmarkedFit classes containing count
history data.
plot.freq logical. Specifies if the count data (pooled across seasons) should be plotted.
plot.seasons logical. Specifies if the count data should be plotted for each season separately.
This argument is only relevant for data collected across more than a single sea-
son.
Details
This function computes a number of summary statistics in data sets used for various N-mixture
models including those of Royle (2004a, b), Dail and Madsen (2011), and Chandler et al. (2011).
countHist can take data frames of the unmarkedFramePCount, unmarkedFrameGPC, unmarkedFrameMPois,
unmarkedFramePCO, unmarkedFrameGMM classes as input. For convenience, the function can also
extract the raw data from model objects of classes unmarkedFitPCount, unmarkedFitGPC, unmarkedFitMPois,
unmarkedFitPCO, and unmarkedFitGMM. Note that different model objects using the same data set
will have identical values.
Value
countHist returns a list with the following components:
countHist 73
count.table.full
a table with the frequency of each observed count.
count.table.seasons
a list of tables with the frequency of each season-specific count.
hist.table.full
a table with the frequency of each count history across the entire sampling pe-
riod.
hist.table.seasons
a list of tables with the frequency of each count history for each primary period
(season).
(common).
no changes observed in season t.
n.visits.season
Author(s)
Marc J. Mazerolle
References
Chandler, R. B., Royle, J. A., King, D. I. (2011) Inference about density and temporary emigration
in unmarked populations. Ecology 92, 1429–1435.
Royle, J. A. (2004a) N-mixture models for estimating population size from spatially replicated
Royle, J. A. (2004b) Generalized estimators of avian abundance from count survey data. Animal
Biodiversity and Conservation 27, 375–386.
See Also
covDiag, detHist, countDist, Nmix.chisq, Nmix.gof.test
74 covDiag
Examples
##modified example from ?pcount

## Not run:
data(mallard)
mallardUMF <- unmarkedFramePCount(mallard.y, siteCovs = mallard.site,
obsCovs = mallard.obs)
countHist(mallardUMF)
##run single season model

fm.mallard <- pcount(~ ivel+ date + I(date^2) ~ length + elev +
forest, mallardUMF, K=30)
countHist(fm.mallard)
}
## End(Not run)
covDiag Compute Covariance Diagnostic for Lambda in N-mixture Models
Description
This function extracts the covariance diagnostic of Dennis et al. (2015) for lambda in N-mixture
models (Royle 2004) of the unmarkedFitPCount class as well as in data frames of the unmarkedFramePcount
class.
Usage
covDiag(object, ...)

## S3 method for class 'unmarkedFramePCount'

Arguments
object an object of class unmarkedFitPCount or unmarkedFramePCount.

covDiag 75
Details
This function extracts the covariance diagnostic developed by Dennis et al. (2015) for lambda in
N-mixture models. Values <= 0 suggest sparse data and potential problems during model fitting.
covDiag can take data frames of the unmarkedFramePcount class as input. For convenience, the
function also takes the repeated count model object as input, extracts the raw data, and computes
the covariance diagnostic. Thus, different models on the same data set will have identical values for
this covariance diagnostic.
Value
covDiag returns a list with the following components:
cov.diag the value of the covariance diagnostic.

message a string indicating whether a warning was issued (i.e., "Warning: lambda is infinite, data too spar
or not (i.e., NULL).
Author(s)
Marc J. Mazerolle
References
Dennis, E. B., Morgan, B. J. T., Ridout, M. S. (2015) Computational aspects of N-mixture models.
Biometrics 71, 237–246.
See Also
modavg, modavgPred, Nmix.chisq, Nmix.gof.test, predictSE, pcount
Examples
##modified example from ?pcount
## Not run:
##Simulate data
set.seed(3)
nSites <- 100
nVisits <- 3
##covariate
x <- rnorm(nSites)
beta0 <- 0
beta1 <- 1
##expected counts
lambda <- exp(beta0 + beta1*x)
N <- rpois(nSites, lambda)
y <- matrix(NA, nSites, nVisits)
p <- c(0.3, 0.6, 0.8)
76 c_hat

}
## Organize data

## Fit model
fm1 <- pcount(~ visit ~ 1, umf, K=50)
covDiag(fm1)
##sparser data
p <- c(0.01, 0.001, 0.01)
}
## Organize data

## Fit model
fm.sparse <- pcount(~ visit ~ 1, umf, K=50)
covDiag(fm.sparse)
}
## End(Not run)
c_hat Estimate Dispersion for Poisson and Binomial GLM’s and GLMM’s
Description
Functions to compute an estimate of c-hat for binomial or Poisson GLM’s and GLMM’s using
different estimators of overdispersion.
Usage
c_hat(mod, method = "pearson", ...)



c_hat 77

Arguments
mod an object of class glm, glmmTMB, merMod, or vglm for which a c-hat estimate is
required.
method this argument defines the estimator used. The default "pearson" uses the Pear-
son chi-square divided by the residual degrees of freedom. Other methods in-
clude "deviance" consisting of the residual deviance divided by the residual
degrees of freedom, "farrington" for the estimator suggested by Farrington
(1996), and "fletcher" for the estimator suggested by Fletcher (2012).
Details
Poisson and binomial GLM’s do not have a parameter for the variance and it is usually held fixed to 1
(i.e., mean = variance). However, one must check whether this assumption is appropriate by estimat-
ing the overdispersion parameter (c-hat). Though one can obtain an estimate of c-hat by dividing the
residual deviance by the residual degrees of freedom (i.e., method = "deviance"), McCullagh and
Nelder (1989) and Venables and Ripley (2002) recommend using Pearson’s chi-square divided by
the residual degrees of freedom (method = "pearson"). An estimator based on Farrington (1996)
is also implemented by the function using the argument method = "farrington". Recent work
by Fletcher (2012) suggests that an alternative estimator performs better than the above-mentioned
methods in the presence of sparse data and is now implemented with method = "fletcher". For
GLMM’s, only the Pearson chi-square estimator of overdispersion is currently implemented.
Note that values of c-hat > 1 indicate overdispersion (variance > mean), but that values much higher
than 1 (i.e., > 4) probably indicate lack-of-fit. In cases of moderate overdispersion, one usually
multiplies the variance-covariance matrix of the estimates by c-hat. As a result, the SE’s of the
estimates are inflated (c-hat is also known as a variance inflation factor).
In model selection, c-hat should be estimated from the global model of the candidate model set and
the same value of c-hat applied to the entire model set. Specifically, a global model is the most
complex model which can be simplified to obtain all the other (nested) models of the set. When
no single global model exists in the set of models considered, such as when sample size does not
allow a complex model, one can estimate c-hat from ’subglobal’ models. Here, ’subglobal’ models
denote models from which only a subset of the models of the candidate set can be derived. In such
cases, one can use the smallest value of c-hat for model selection (Burnham and Anderson 2002).
Note that c-hat counts as an additional parameter estimated and should be added to K. All functions
in package AICcmodavg automatically add 1 when the c.hat argument > 1 and apply the same
value of c-hat for the entire model set. When c.hat > 1, functions compute quasi-likelihood
information criteria (either QAICc or QAIC, depending on the value of the second.ord argument)
by scaling the log-likelihood of the model by c.hat. The value of c.hat can influence the ranking
of the models: as c-hat increases, QAIC or QAICc will favor models with fewer parameters. As
an additional check against this potential problem, one can create several model selection tables by
incrementing values of c-hat to assess the model selection uncertainty. If ranking changes little up
to the c-hat value observed, one can be confident in making inference.
78 c_hat
In cases of underdispersion (c-hat < 1), it is recommended to keep the value of c.hat to 1. How-
ever, note that values of c-hat « 1 can also indicate lack-of-fit and that an alternative model (and
distribution) should be investigated.
Note that c_hat only supports the estimation of c-hat for binomial models with trials > 1 (i.e.,
success/trial or cbind(success, failure) syntax) or with Poisson GLM’s or GLMM’s.
Value
c_hat returns an object of class c_hat with the estimated c-hat value and an attribute for the type
of estimator used.
Author(s)
Marc J. Mazerolle
References
New York.
Farrington, C. P. (1996) On assessing goodness of fit of generalized linear models to sparse data.
Journal of the Royal Statistical Society B 58, 349–360.
Fletcher, D. J. (2012) Estimating overdispersion when fitting a generalized linear model to sparse
data. Biometrika 99, 230–237.
Hall: New York.
Venables, W. N., Ripley, B. D. (2002) Modern Applied Statistics with S. Second edition. Springer:
New York.
See Also
AICc, confset, evidence, modavg, importance, modavgPred, mb.gof.test, Nmix.gof.test,
anovaOD, summaryOD
Examples
#binomial glm example
set.seed(seed = 10)
resp <- rbinom(n = 60, size = 1, prob = 0.5)
set.seed(seed = 10)
treat <- as.factor(sample(c(rep(x = "m", times = 30), rep(x = "f",
times = 30))))
age <- as.factor(c(rep("young", 20), rep("med", 20), rep("old", 20)))
detHist 79
#each invidual has its own response (n = 1)

mod1 <- glm(resp ~ treat + age, family = binomial)
## Not run:
c_hat(mod1) #gives an error because model not appropriate for
##computation of c-hat
## End(Not run)
##computing table to summarize successes

table(resp, treat, age)
dat2 <- as.data.frame(table(resp, treat, age)) #not quite what we need
data2 <- data.frame(success = c(9, 4, 2, 3, 5, 2),
sex = c("f", "m", "f", "m", "f", "m"),
age = c("med", "med", "old", "old", "young",
"young"), total = c(13, 7, 10, 10, 7, 13))
data2$prop <- data2$success/data2$total
data2$fail <- data2$total - data2$success
##run model with success/total syntax using weights argument

mod2 <- glm(prop ~ sex + age, family = binomial, weights = total,
data = data2)
c_hat(mod2)
##run model with other syntax cbind(success, fail)

mod3 <- glm(cbind(success, fail) ~ sex + age, family = binomial,
data = data2)
c_hat(mod3)
detHist Compute Summary Statistics from Detection Histories
Description
This function extracts various summary statistics from detection history data of various unmarkedFrame
and unmarkedFit classes.
Usage
detHist(object, ...)



80 detHist

## S3 method for class 'unmarkedFitOccuMulti'

## S3 method for class 'unmarkedFrameOccu'

## S3 method for class 'unmarkedFrameOccuFP'

## S3 method for class 'unmarkedMultFrame'

## S3 method for class 'unmarkedFrameOccuMulti'

Arguments
object an object of various unmarkedFrame or unmarkedFit classes containing detec-

tion history data.
Details
This function computes a number of summary statistics in data sets used for single-season occu-
pancy models (MacKenzie et al. 2002), dynamic occupancy models (MacKenzie et al. 2003),
Royle-Nichols models (Royle and Nichols 2003), false-positive occupancy models (Royle and Link
2006, Miller et al. 2011), and multispecies occupancy models (Rota et al. 2016.
detHist can take data frames of the unmarkedFrameOccu, unmarkedFrameOccuFP, unmarkedMultFrame,
and unmarkedFrameOccuMulti classes as input. For convenience, the function can also extract the
raw data from model objects of classes unmarkedFitColExt, unmarkedFitOccu, unmarkedFitOccuFP,
unmarkedFitOccuRN, and unmarkedFrameOccuMulti. Note that different model objects using the
same data set will have identical values.
Value
For objects of classes unmarkedFitOccu, unmarkedFitOccuRN, unmarkedFitOccuRN, unmarkedFitColExt,

unmarkedFrameOccu, unmarkedFrameOccuFP, and unmarkedMultFrame, detHist returns a list
with the following components:
hist.table.full
a table with the frequency of each observed detection history.
hist.table.seasons
a list of tables with the frequency of each season-specific detection history.
detHist 81
(common). For multispecies data, out.freqs presents for each species the num-
ber of sites sampled and the number of sites with at least one detection.
no changes observed in season t. For multispecies data, out.props presents the
proportion of sites with a least one detection for each species.
n.visits.season
n.species the number of species in the data set.
For objects of classes unmarkedFitOccuMulti and unmarkedFrameOccuMulti, detHist returns a
list with the following components:
hist.table.full
a table with the frequency of each observed detection history. The species are
coded with letters and follow the same order of presentation as in the other parts
of the output.
hist.table.species
a list of tables with the frequency of each species-specific detection history.
The last element of hist.table.species features the number of sites with co-
occurrence of the different species (coOcc).
out.freqs a matrix where the rows correspond to each species and where columns consist
of the number of sites sampled during the season (sampled) and the number of
sites with at least one detection (detected).
out.props a matrix where the rows correspond to each species and where columns con-
sist of the proportion of sites with at least one detection during the season
(naive.occ).
n.visits.season
n.species the number of species in the data set.
Author(s)
Marc J. Mazerolle
82 detHist
References
2248–2255.
84, 2200–2207.
Miller, D. A. W., Nichols, J. D., McClintock, B. T., Campbell Grant, E. H., Bailey, L. L. (2011)
Improving occupancy estimation when two types of observational error occur: non-detection and
species misidentification. Ecology 92, 1422–1428.
Rota, C. T., Ferreira, M. A. R., Kays, R. W., Forrester, T. D., Kalies, E. L., McShea, W. J., Parsons,
A. W., Millspaugh, J. J. (2016) A multispecies occupancy model for two or more interacting species.
Methods in Ecology and Evolution 7, 1164–1173.
Royle, J. A., Link, W. A. (2006) Generalized site occupancy models allowing for false positive and
false negative errors. Ecology 87, 835–841.
Royle, J. A., Nichols, J. D. (2003) Estimating abundance from repeated presence-absence data or
point counts. Ecology 84, 777–790.
See Also
covDiag, countHist, countDist, mb.chisq, mb.gof.test,
Examples
##data from Mazerolle (2015)
## Not run:
data(bullfrog)
##detection data


detHist(bfrog)
##run model
fm <- occu(~ 1 ~ 1, data = bfrog)
detHist(fm)
}
## End(Not run)
DIC 83
DIC Computing DIC
Description
Functions to extract deviance information criterion (DIC).
Usage
DIC(mod, return.pD = FALSE, ...)
## S3 method for class 'bugs'

## S3 method for class 'rjags'

## S3 method for class 'jagsUI'

Arguments
mod an object of class bugs, rjags, or jagsUI containing the output of a model.
return.pD logical. If FALSE, the function returns the DIC. If TRUE, the function returns the
effective number of estimated parameters (pD) for a given model.
Details
DIC is implemented for bugs, rjags, and jagsUI classes. The function extracts the deviance infor-
mation criterion (DIC, Spiegelhalter et al. 2002) or the effective number of parameters (pD).
Value
DIC the DIC or pD depending on the values of the arguments.
Note
The actual DIC values are not really interesting in themselves, as they depend directly on the data,
parameters estimated, and likelihood function. Furthermore, a single value does not tell much about
model fit. Information criteria become relevant when compared to Yone another for a given data set
and set of candidate models. Model selection with hierarchical models is problematic as the classic
DIC is not appropriate for such types of models (Millar 2009).
Author(s)
Marc J. Mazerolle
84 DIC
References
Millar, R. B. (2009) Comparison of hierarchical Bayesian models for overdispersed count data using
DIC and Bayes’ factors. Biometrics, 65, 962–969.
Spiegelhalter, D. J., Best, N. G., Carlin, B. P., van der Linde, A. (2002). Bayesian measures of
complexity and fit. Journal of the Royal Statistical Society, Series B 64, 583–639.
See Also
AICcCustom, aictab, dictab, confset, evidence
Examples
##from ?jags example in R2jags package
## Not run:
require(R2jags)
##example model file
model.file <- system.file(package="R2jags", "model", "schools.txt")
file.show(model.file)
##data
J <- 8.0
y <- c(28.4,7.9,-2.8,6.8,-0.6,0.6,18.0,12.2)
sd <- c(14.9,10.2,16.3,11.0,9.4,11.4,10.4,17.6)
##arrange data in list

jags.data <- list (J = J, y = y, sd = sd)
##initial values
jags.inits <- function(){
list(theta=rnorm(J, 0, 100), mu=rnorm(1, 0, 100),
sigma=runif(1, 0, 100))
}
##parameters to be monitored
jags.parameters <- c("theta", "mu", "sigma")
##run model
schools.sim <- jags(data = jags.data, inits = jags.inits,
parameters = jags.parameters,
model.file = model.file,
n.chains = 3, n.iter = 10)
##note that n.iter should be higher
##extract DIC
DIC(schools.sim)
##extract pD
DIC(schools.sim, return.pD = TRUE)
detach(package:R2jags)
## End(Not run)
dictab 85
dictab Create Model Selection Tables from Bayesian Analyses
Description
This function creates a model selection table based on the deviance information criterion (DIC).
The table ranks the models based on the DIC and also provides delta DIC and DIC weights. dictab
selects the appropriate function to create the model selection table based on the object class. The
current version works with objects of bugs, rjags, jagsUI classes.
Usage
dictab(cand.set, modnames = NULL, sort = TRUE, ...)
## S3 method for class 'AICbugs'

## S3 method for class 'AICrjags'

## S3 method for class 'AICjagsUI'

Arguments
models.
sort logical. If TRUE, the model selection table is ranked according to the DIC values.
Details
dictab internally creates a new class for the cand.set list of candidate models, according to the
contents of the list. The current function is implemented for bugs, jags, jagsUI classes. The
function constructs a model selection table based on the DIC (Spiegelhalter et al. 2002). Note that
DIC might not be appropriate to select among a set of hierarchical models and that modifications to
the information criterion have been proposed (Millar 2009).
Value
dictab creates an object of class dictab with the following components:

86 dictab
pD the effective number of estimated parameters for each model.

DIC the deviance information criterion for each model.
Delta_DIC the delta DIC of each model, measuring the difference in DIC between each
model and the top-ranked model.
probabilities.
DICWt the DIC weights, sensu Burnham and Anderson (2002) and Anderson (2008).
These measures indicate the level of support (i.e., weight of evidence) in favor
of any given model being the most parsimonious among the candidate model
set.
Cum.Wt the cumulative DIC weights. These are only meaningful if results in table are
sorted in decreasing order of DIC weights (i.e., sort = TRUE).
Deviance the deviance of each model.
Author(s)
Marc J. Mazerolle
References
New York.
Spiegelhalter, D. J., Best, N. G., Carlin, B. P., van der Linde, A. (2002). Bayesian measures of
complexity and fit. Journal of the Royal Statistical Society, Series B 64, 583–639.
See Also
aictabCustom, aictab, confset, DIC, evidence
Examples
##from ?jags example in R2jags package
## Not run:
require(R2jags)
model.file <- system.file(package="R2jags", "model", "schools.txt")
file.show(model.file)
##data
J <- 8.0
y <- c(28.4,7.9,-2.8,6.8,-0.6,0.6,18.0,12.2)
sd <- c(14.9,10.2,16.3,11.0,9.4,11.4,10.4,17.6)
jags.data <- list (J = J, y = y, sd = sd)

dry.frog 87
jags.inits <- function(){

list(theta=rnorm(J, 0, 100), mu=rnorm(1, 0, 100),
sigma=runif(1, 0, 100))
}
jags.parameters <- c("theta", "mu", "sigma")
##run model
schools.sim <- jags(data = jags.data, inits = jags.inits,
parameters = jags.parameters,
model.file = model.file,
n.chains = 3, n.iter = 10)
#note that n.iter should be higher
##set up in list
Cand.mods <- list(schools.sim)
Model.names <- "hierarchical model"
##other models can be added to Cand.mods
##to compare them to the top model
##model selection table

dictab(cand.set = Cand.mods, modnames = Model.names)
detach(package:R2jags)
## End(Not run)
dry.frog Frog Dehydration Experiment on Three Substrate Types
Description
This is a data set modified from Mazerolle and Desrochers (2005) on the mass lost by frogs after
spending two hours on one of three substrates that are encountered in some landscape types.
Usage
data(dry.frog)
Format
Individual a numeric identifier unique to each individual.
Species a factor with levels Racla.
Shade a numeric vector, either 1 (shade) or 0 (no shade).
SVL the snout-vent length of the individual.
Substrate the substrate type, a factor with levels PEAT, SOIL, and SPHAGNUM.
Initial_mass the initial mass of individuals.
Mass_lost the mass lost in g.
88 evidence
Airtemp the air temperature in degrees C.

Wind_cat the wind intensity, either 0 (no wind), 1 (low wind), 2 (moderate wind), or 3 (strong
wind).
Cloud cloud cover expressed as a percentage.
cent_Initial_mass centered inital mass.
Initial_mass2 initial mass squared.
cent_Air centered air temperature.
Perc.cloud proportion of cloud cover
Wind wind intensity, either 1 (no or low wind) or 1 (moderate to strong wind).
log_Mass_lost log of mass lost.
Details
Note that the original analysis in Mazerolle and Desrochers (2005) consisted of generalized estimat-
ing equations for three mass measurements: mass at time 0, 1 hour, and 2 hours following exposure
on the substrate.
Source
Mazerolle, M. J., Desrochers, A. (2005) Landscape resistance to frog movements. Canadian Jour-
nal of Zoology 83, 455–464.
Examples
data(dry.frog)
## maybe str(dry.frog) ; plot(dry.frog) ...
evidence Compute Evidence Ratio Between Two Models
Description
This function compares two models of a candidate model set based on their evidence ratio (i.e.,
ratio of model weights). The default computes the evidence ratio of the model weights between the
top-ranked model and the second-ranked model. You must supply a model selection table of class
aictab, bictab, boot.wt, dictab, ictab as the first argument.
Usage
evidence(aic.table, model.high = "top", model.low = "second.ranked")

evidence 89
Arguments
aic.table a model selection table of class aictab such as that produced by aictab or of
classes bictab, boot.wt, dictab, or ictab. The table may be sorted or not, as
the function sorts the table internally.
model.high the top-ranked model (default), or alternatively, the name of another model as it
appears in the model selection table.
model.low the second-ranked model (default), or alternatively, the name of a lower-ranked
model such as it appears in the model selection table.
Details
The default compares the model weights of the top-ranked model to the second-ranked model in the
candidate model set. The evidence ratio can be interpreted as the number of times a given model
is more parsimonious than a lower-ranked model. If one desires an evidence ratio that does not
involve a comparison with the top-ranking model, the label of the required model must be specified
in the model.high argument as it appears in the model selection table.
Value
evidence produces an object of class evidence with the following components:
Model.high the model specified in model.high.

Model.low the model specified in model.low.
Ev.ratio the evidence ratio between the two models compared.
Author(s)
Marc J. Mazerolle
References
See Also
AICc, aictab, bictab, c_hat, confset, importance, modavg, modavgShrink, modavgPred
Examples
##run example from Burnham and Anderson (2002, p. 183) with two
##non-nested models
data(pine)
Cand.set <- list( )
Cand.set[[1]] <- lm(y ~ x, data = pine)
Cand.set[[2]] <- lm(y ~ z, data = pine)
##assign model names

Modnames <- c("raw density", "density corrected for resin content")
90 evidence
##compute model selection table

aicctable.out <- aictab(cand.set = Cand.set, modnames = Modnames)

evidence(aic.table = aicctable.out, model.low = "raw density")
evidence(aic.table = aicctable.out) #gives the same answer
print(evidence(aic.table = aicctable.out, model.low = "raw density"),
digits = 4)
##example with bictab

## Not run:
##compute model selection table
bictable.out <- bictab(cand.set = Cand.set, modnames = Modnames)
evidence(bictable.out, model.low = "raw density")
## End(Not run)
##run models for the Orthodont data set in nlme package

## Not run:
require(nlme)
##set up candidate model list

Cand.models <- list()
Cand.models[[1]] <- lme(distance ~ age, data = Orthodont, method = "ML")
##random is ~ age | Subject
random = ~ 1, method = "ML")
Cand.models[[3]] <- lme(distance ~ 1, data = Orthodont, random = ~ 1,
method = "ML")

##compute AICc table

aic.table.1 <- aictab(cand.set = Cand.models, modnames = Modnames,
second.ord = TRUE)
##compute evidence ratio between best model and second-ranked model

evidence(aic.table = aic.table.1)
##compute the same value but from an unsorted model selection table
evidence(aic.table = aictab(cand.set = Cand.models,
modnames = Modnames, second.ord = TRUE, sort = FALSE))
##compute evidence ratio between second-best model and third-ranked

##model
evidence(aic.table = aic.table.1, model.high = "mod1",
model.low = "mod3")
extractCN 91
## End(Not run)
extractCN Compute Condition Number
Description
This function computes the condition number for models of unmarkedFit classes as the ratio of the
largest eigenvalue of the Hessian matrix to the smallest eigenvalue of the Hessian matrix.
Usage
extractCN(mod, method = "svd", ...)

extractCN(mod, method = "svd", ...)
Arguments
mod a model of one the unmarkedFit classes for which a condition number is re-
quested.
method specifies the method used to extract the singular values or eigenvalues from
the Hessian matrix using singular value decomposition (method = "svd") or
eigenvalue decomposition (method = "eigen").
Details
The condition number (κ) is a measure of the transfer of error to the solution in response to small
changes in the input (Cheney and Kincaid 2008). In this implementation, the condition number is
computed on the Hessian matrix of models of unmarkedFit classes from the optim results stored
in the model object. The condition number is defined as the ratio of the largest to the smallest non-
negative singular values of a given matrix (Cline et al. 1979, Dixon 1983). In the special case of
positive semi-definite matrices, the singular values are equal to the eigenvalues (Ruhe 1975).
Large values of the condition number may indicate problems in estimating parameters or their
variance (ill-conditioning), possibly due to a model having too many parameters for the given data
set. Cheney and Ward (2008) suggest using the log10 (κ) of the condition number as a crude estimate
of the number of digits of precision lost.
92 extractCN
Value
extractCN returns a list of class extractCN with the following components:
CN the condition number (κ) of the model.

log10 the log base 10 of the condition number.
method the method used to extract the singular values or eigenvalues.
Author(s)
Marc J. Mazerolle
References
Cheney, W., Kincaid, D. (2008) Numerical mathematics and computing. Sixth edition. Thomson
Brooks/Cole: Belmont.
Cline, A. K., Moler, C. B., Stewart, G. W., Wilkinson, J. H. (1979) An estimate for the condition
number of a matrix. SIAM Journal on Numerical Analysis 16, 368–375.
Dixon, J. D. (1983) Estimating extremal eigenvalues and condition numbers of matrices. SIAM
Journal on Numerical Analysis 20, 812–814.
Ruhe, A. (1975) On the closeness of eigenvalues and singular values for almost normal matrices.
Linear Algebra and its Applications 11, 87–94.
See Also
c_hat, mb.gof.test, Nmix.gof.test, parboot, kappa, rcond
Examples
##N-mixture model example modified from ?pcount
## Not run:
require(unmarked)
##single season
data(mallard)
##run model
fm.mallard <- pcount(~ ivel+ date + I(date^2) ~ length + elev + forest,
mallardUMF, K=30)
##compute condition number

extractCN(fm.mallard)
##compare against 'kappa'

kappa(fm.mallard@opt$hessian, exact = TRUE)
## End(Not run)
extractLL 93
extractLL Extract Log-Likelihood of Model
Description
This function extracts the log-likelihood from an object of coxme, coxph, lmekin, maxlikeFit,
vglm, or various unmarkedFit classes.
Usage
extractLL(mod, ...)

extractLL(mod, type = "Integrated", ...)

extractLL(mod, ...)

extractLL(mod, ...)

extractLL(mod, ...)

extractLL(mod, ...)

extractLL(mod, ...)
Arguments
mod an object of coxme, coxph, lmekin, maxlikeFit, vglm, or unmarkedFit class

resulting from the fit of distsamp, gdistsamp, gmultmix, multinomPois, gpcount,
occu, occuRN, colext, pcount, or pcountOpen.
type a character string indicating whether the integrated partial likelihood ("Inte-
grated") or penalized likelihood ("Penalized") is to be used for a coxme object.
Details
This utility function extracts the information from a coxme, coxph, lmekin, maxlikeFit, vglm, or
unmarkedFit object resulting from distsamp, gdistsamp, gmultmix, multinomPois, gpcount,
occu, occuRN, colext, pcount, or pcountOpen.
94 extractSE
Value
These functions return the value of the log-likelihood of the model and associated degrees of free-
dom.
Author(s)
Marc J. Mazerolle
See Also
AICc, aictab, coxme, coxph, lmekin, maxlike, distsamp, gdistsamp, occu, occuRN, colext,
pcount, pcountOpen
Examples
## Not run:
require(unmarked)
##single season
data(frogs)

obsNum(pferUMF)))
##run model set

extractLL(fm1)
## End(Not run)
extractSE Extract SE of Fixed Effects
Description
This function extracts the standard errors (SE) of the fixed effects of a mixed model fit with coxme,
glmer, lmer, lmerModLmerTest, and lmekin and adds the appropriate labels.
extractSE 95
Usage
extractSE(mod, ...)

extractSE(mod, ...)

extractSE(mod, ...)

extractSE(mod, ...)

extractSE(mod, ...)

extractSE(mod, ...)
Arguments
mod an object of coxme, lmekin, mer, merMod, or lmerModTest class.
Details
These extractor functions use vcov.coxme, vcov.lmekin, vcov.mer, and vcov.merMod. Some of
these functions are called by modavg and modavgShrink, depending on the class of the objects.
Value
Returns the SE’s of the fixed effects with the appropriate labels for each.
Author(s)
Marc J. Mazerolle
See Also
modavg, glmer, lmer, coxme, lmekin
Examples
##modified example from ?glmer
## Not run:
if(require(lme4)) {
##create proportion of incidence
cbpp$prop <- cbpp$incidence/cbpp$size
gm1 <- glmer(prop ~ period + (1 | herd), family = binomial,
96 extractX
weights = size, data = cbpp)

##print summary
summary(gm1)
##extract variance-covariance matrix of fixed effects
vcov(gm1)
##extract SE's of fixed effects - no labels
sqrt(diag(vcov(gm1))) #no labels
extractSE(gm1) #with labels
detach(package:lme4)
}
## End(Not run)
extractX Extract Predictors from Candidate Model List
Description
This function extracts the predictors used in candidate models. The function is currently imple-
mented for glm, glmmTMB, gls, lm, lme, merMod, lmerModLmerTest, rlm, survreg object classes
that are stored in a list as well as various models of unmarkedFit classes.
Usage
extractX(cand.set, ...)








extractX 97




extractX(cand.set,
parm.type = NULL, ...)

extractX(cand.set,

extractX(cand.set,

extractX(cand.set,

extractX(cand.set,

extractX(cand.set,

extractX(cand.set,

extractX(cand.set,

extractX(cand.set,

98 extractX
extractX(cand.set,

extractX(cand.set,

extractX(cand.set,
Arguments
parm.type this argument specifies the parameter type on which the effect size will be com-
puted and is only relevant for models of unmarkedFitOccu, unmarkedFitColExt,
unmarkedFitOccuFP, unmarkedFitOccuMulti, unmarkedFitOccuRN, unmarkedFitMPois,
unmarkedFitPCount, unmarkedFitPCO, unmarkedFitDS, unmarkedFitGDS, unmarkedFitGMM,
and unmarkedFitGPC classes. The character strings supported vary with the
type of model fitted. For unmarkedFitOccu and unmarkedFitOccuMulti ob-
jects, either psi or detect can be supplied to indicate whether the parameter
is on occupancy or detectability, respectively. For unmarkedFitColExt, pos-
sible values are psi, gamma, epsilon, and detect, for parameters on occu-
pancy in the inital year, colonization, extinction, and detectability, respectively.
For unmarkedFitOccuFP objects, one can specify psi, detect, falsepos, and
certain, for occupancy, detectability, probability of assigning false-positives,
and probability detections are certain, respectively. For unmarkedFitOccuRN
objects, either lambda or detect can be entered for abundance and detectability
parameters, respectively. For unmarkedFitPCount and unmarkedFitMPois ob-
jects, lambda or detect denote parameters on abundance and detectability, re-
spectively. For unmarkedFitPCO objects, one can enter lambda, gamma, omega,
iota, or detect, to specify parameters on abundance, recruitment, apparent sur-
vival, immigration, and detectability, respectively. For unmarkedFitDS objects,
lambda and detect are supported. For unmarkedFitGDS, lambda, phi, and
detect denote abundance, availability, and detection probability, respectively.
For unmarkedFitGMM and unmarkedFitGPC objects, lambda, phi, and detect
denote abundance, availability, and detectability, respectively.
Details
The candidate models must be stored in a list. The results of extractX are useful in preparing a
newdata data frame to use in computing model-averaged predictions with modavgPred or differ-
ences between groups with modavgEffect (Burnham and Anderson 2002, Anderson 2008, Burn-
ham et al. 2011).
Value
extractX returns an object of class extractX with the following components:
extractX 99
predictors a character vector of the names of the predictors included in the model, exclud-
ing the intercept term.
data a data frame or, in the case of unmarkedFit objects, a list of data frames (e.g.,
obsCovs, siteCovs, yearlySiteCovs).
Author(s)
Marc J. Mazerolle
References
New York.
Burnham, K. P., Anderson, D. R., Huyvaert, K. P. (2011) AIC model selection and multimodel
inference in behaviorial ecology: some background, observations and comparisons. Behavioral
Ecology and Sociobiology 65, 23–25.
Pinheiro, J. C., Bates, D. M. (2000). Mixed-effects Models in S and S-PLUS. Springer Verlag: New
York.
See Also
extractCN, extractSE, modavgPred, modavgCustom, modavgEffect, predict, predictSE
Examples
##example from subset of models in Table 1 in Mazerolle (2006)
data(dry.frog)

data = dry.frog)
100 extractX
##assign names
names(Cand.models) <- paste(1:length(Cand.models))
##extract predictors from candidate model set

orig.data <- extractX(cand.set = Cand.models)
orig.data
str(orig.data)
## Not run:
##model-averaged prediction with original variables
modavgPred(Cand.models, newdata = orig.data$data)
## End(Not run)
##example of model-averaged predictions from N-mixture model (e.g., Royle 2004)

##modified from ?pcount
##each variable appears twice on lambda in the models
## Not run:
require(unmarked)
data(mallard)
##set up models so that each variable on abundance appears twice
fm.mall.one <- pcount(~ ivel + date ~ length + forest, mallardUMF,
K = 30)
fm.mall.two <- pcount(~ ivel + date ~ elev + forest, mallardUMF,
K = 30)
fm.mall.three <- pcount(~ ivel + date ~ length + elev, mallardUMF,
K = 30)
fm.mall.four <- pcount(~ ivel + date ~ 1, mallardUMF, K = 30)
##model list
Cands <- list(fm.mall.one, fm.mall.two, fm.mall.three, fm.mall.four)
names(Cands) <- c("length + forest", "elev + forest", "length + elev",
"null")
##extract predictors on lambda

lam.dat <- extractX(cand.set = Cands, parm.type = "lambda")
lam.dat
str(lam.dat)
##extract predictors on detectability

extractX(cand.set = Cands, parm.type = "detect")
##model-averaged predictions on lambda

##extract data
siteCovs <- lam.dat$data$siteCovs
##create vector of forest values
forest <- seq(min(siteCovs$forest),
max(siteCovs$forest),
length.out = 40)
dframe <- data.frame(forest = forest,
extractX 101
length = mean(siteCovs$length),
elev = mean(siteCovs$elev))
modavgPred(Cands, parm.type = "lambda",
newdata = dframe)
## End(Not run)
##example of model-averaged abundance from distance model

## Not run:
require(unmarked)
data(linetran) #example from ?distsamp

dist.breaks = c(0, 5, 10, 15, 20),
unitsIn = "m")
})
## Half-normal detection function. Density output (log scale). No covariates.

## Halfnormal. Covariates affecting both density and and detection.

fm2 <- distsamp(~area + habitat ~ habitat, ltUMF)
## Hazard function. Covariates affecting both density and and detection.

fm3 <- distsamp(~area + habitat ~ habitat, ltUMF, keyfun="hazard")
##assemble model list

Cands <- list(fm1, fm2, fm3)
##model-average predictions on abundance

extractX(cand.set = Cands, parm.type = "lambda")
## End(Not run)
##example using Orthodont data set from Pinheiro and Bates (2000)
## Not run:
require(nlme)

m1 <- gls(distance ~ age, correlation = corCompSymm(value = 0.5, form = ~ 1 | Subject),
m2 <- gls(distance ~ 1, correlation = corCompSymm(value = 0.5, form = ~ 1 | Subject),

102 fam.link.mer
##assemble in list
Cand.models <- list("age effect" = m1, "null model" = m2)
##model-averaged predictions
extractX(cand.set = Cand.models)
## End(Not run)
fam.link.mer Extract Distribution Family and Link Function
Description
This function extracts the distribution family and link function of a generalized linear mixed model
fit with glmer or lmer.
Usage
fam.link.mer(mod)
Arguments
mod an object of mer or merMod class resulting from the fit of glmer or lmer.
Details
This utility function extracts the information from an mer or merMod object resulting from glmer or
lmer. The function is called by modavg, modavgEffect, modavgPred, and predictSE.
Value
fam.link.mer returns a list with the following components:
family the family of the distribution of the model.

link the link function of the model.
supp.link a character value indicating whether the link function used is supported by
predictSE and modavgPred.
Author(s)
Marc J. Mazerolle
See Also
modavg, modavgPred, predictSE, glmer, lmer
fat 103
Examples
##modified example from ?glmer
## Not run:
if(require(lme4)){
##create proportion of incidence
cbpp$prop <- cbpp$incidence/cbpp$size
gm1 <- glmer(prop ~ period + (1 | herd), family = binomial,
weights = size, data = cbpp)
fam.link.mer(gm1)
gm2 <- glmer(prop ~ period + (1 | herd),
family = binomial(link = "cloglog"), weights = size,
data = cbpp)
fam.link.mer(gm2)
}
## End(Not run)
##example with linear mixed model with Orthodont data from

##Pinheiro and Bates (2000)
## Not run:
data(Orthodont, package = "nlme")
m1 <- lmer(distance ~ Sex + (1 | Subject), data = Orthodont,
REML = FALSE)
fam.link.mer(m1)
m2 <- glmer(distance ~ Sex + (1 | Subject),
family = gaussian(link = "log"), data = Orthodont,
REML = FALSE)
fam.link.mer(m2)
detach(package:lme4)
## End(Not run)
fat Fat Data and Body Measurements
Description
This data set illustrates the relationship between body measurements and body fat in 252 males
aged between 21 and 81 years.
Usage
data(fat)
Format
Obs observation number.
104 fat
Perc.body.fat.Brozek percent body fat using Brozek’s equation, i.e., 457/Density − 414.2.
Perc.body.fat.Siri percent body fat using Siri’s equation, i.e., 495/Density − 450.
Density density ( cm
g
3 ).
Age age (years).

Weight weight (lbs).
Height height (inches).
Adiposity.index adiposity index computed as W eight/Height2 ( m
kg
2 ).
Fat.free.weight fat free weight computed as (1 − Brozek 0 spercentbodyf at) ∗ W eight (lbs).
Neck.circ neck circumference (cm).
Chest.circ chest circumference (cm).
Abdomen.circ abdomen circumference (cm) measured at the umbilicus and level with the iliac
crest.
Hip.circ hip circumference (cm).
Thigh.circ thigh circumference (cm).
Knee.circ knee circumference (cm).
Ankle.circ ankle circumference (cm).
Biceps.circ extended biceps circumference (cm).
Forearm.circ forearm circumference (cm).
Wrist.circ wrist circumference (cm).
3
inv.Density inverse of density ( cm
g ).
z1 log of weight divided by log of height (allometric measure).
z2 abdomen circumference divided by chest circumference (beer gut factor).
(
z3 index based on knee, wrist, and ankle circumference relative to height ( (Knee.circ∗W rist.circ∗Ankle.circ)
Height
1/3)
).
z4 fleshiness index based on biceps, thigh, forearm, knee, wrist, and ankle circumference ( Biceps.circ∗T high.circ∗F orearm.circ
Knee.circ∗W rist.circ∗Ankle.circ
z5 age standardized to zero mean and unit variance.
z6 square of standardized age.
Details
Burnham and Anderson (2002, p. 268) use this data set to show model selection uncertainty in the
context of all possible combinations of explanatory variables. The data are originally from Penrose
et al. (1985) who used only the first 143 cases of the 252 observations in the data set. Johnson
(1996) later used these data as an example of multiple regression. Note that observation number 42
originally had an erroneous height of 29.5 inches and that this value was changed to 69.5 inches.
Burnham and Anderson (2002, p. 274) created six indices based on the original measurements (i.e.,
z1 – z6). Although Burnham and Anderson (2002) indicate that the fleshiness index (z4) involved
the cubic root in the equation, the result table for the full model on p. 276 suggests that the index
did not include the cubic root for z4. The latter is the version of z4 used in the data set here.
gpa 105
Source
Johnson, J. W. (1996). Fitting percentage of body fat to simple body measurements. Journal of
Statistics Education 4 [online].
Penrose, K., Nelson, A., Fisher, A. (1985) Generalized body composition prediction equation for
men using simple measurement techniques Medicine and Science in Sports and Exercise 17, 189.
Examples
data(fat)
str(fat)
gpa GPA Data and Standardized Test Scores
Description
This data set features the first-year college GPA and four standardized tests conducted before ma-
triculation.
Usage
data(gpa)
Format
gpa.y first-year GPA.
sat.math.x1 SAT math score.
sat.verb.x2 SAT verbal score.
hs.math.x3 high school math score.
hs.engl.x4 high school English score.
Details
Burnham and Anderson (2002, p. 225) use this data set originally from Graybill and Iyer (1994) to
show model selection for all subsets regression.
Source
Graybill, F. A., Iyer, H. K. (1994) Regression analysis: concepts and applications. Duxbury Press:
Belmont.
106 ictab
Examples
data(gpa)
str(gpa)
ictab Create Model Selection Tables from User-supplied Information Crite-

rion
Description
This function creates a model selection table from information criterion values supplied by the user.
The table ranks the models based on the values of the information criterion and also displays delta
values and information criterion weights.
Usage
ictab(ic, K, modnames = NULL, sort = TRUE, ic.name = NULL)
Arguments
ic a vector of information criterion values for each model in the candidate model
set.
modnames a character vector of model names to identify each model in the model selection
table. If NULL, generic names (e.g., Mod1, Mod2) are supplied in the table in the
same order as the information criterion values.
sort logical. If TRUE, the model selection table is ranked according to the values of
the information criterion.
ic.name a character string denoting the name of the information criterion input by the
user. This character string will appear in certain column labels of the model
selection table.
Details
ictab constructs a model selection table based on the information criterion values supplied by
the user. This function is most useful for information criterion other than AIC, AICc, QAIC, and
QAICc (e.g., WAIC: Watanabe 2010) or for classes not supported by aictab or bictab.
Value
ictab creates an object of class ictab with the following components:

ictab 107
IC the values of the information criterion input by the user. If a value for ic.name
is provided, the table modifies the labels of the table.
Delta_IC the delta information criterion component comparing each model to the top-
ranked model.
probabilities.
ICWt the information criterion weights, also termed "model probabilities" sensu Burn-
ham and Anderson (2002) and Anderson (2008). These measures indicate the
level of support (i.e., weight of evidence) in favor of any given model being the
most parsimonious among the candidate model set.
Cum.Wt the cumulative information criterion weights. These are only meaningful if re-
sults in table are sorted in decreasing order of Akaike weights (i.e., sort = TRUE).
Author(s)
Marc J. Mazerolle
References
New York.
Watanabe, S. (2010) Asymptotic equivalence of Bayes cross validation and widely applicable infor-
mation criterion in singular learning theory. Journal of Machine Learning Research 11, 3571–3594.
See Also
aictabCustom, confset, evidence, modavgCustom, modavgIC
Examples
##WAIC values
waic <- c(105.74, 107.36, 108.24, 100.57)
effK <- c(7.45, 5.61, 6.14, 6.05)
##generate WAIC table

ictab(ic = waic, K = effK, modnames = Modnames,
sort = TRUE, ic.name = "WAIC")
108 importance
importance Compute Importance Values of Variable
Description
This function calculates the relative importance of variables (w+) based on the sum of Akaike
weights (model probabilities) of the models that include the variable. Note that this measure of
evidence is only appropriate when the variable appears in the same number of models as those that
do not include the variable.
Usage
importance(cand.set, parm, modnames = NULL, second.ord = TRUE,
nobs = NULL, ...)

importance(cand.set, parm, modnames = NULL,
second.ord = TRUE, nobs = NULL, ...)




## S3 method for class 'AICclogit.coxph'



importance 109

second.ord = TRUE, nobs = NULL, c.hat = 1, ...)











110 importance





second.ord = TRUE, nobs = NULL, c.hat = 1, parm.type = NULL, ...)







importance 111






Arguments
parm the parameter of interest for which a measure of relative importance is required.
models.
rion (i.e., AICc).
value of nobs.
mixture models (Royle 2004, Dail and Madsen 2011). If c.hat > 1, importance
112 importance

unmarkedFitOccuFP, unmarkedFitOccuRN, unmarkedFitMPois, unmarkedFitGPC,
unmarkedFitPCount, unmarkedFitPCO, unmarkedFitDS, unmarkedFitGDS, unmarkedFitGMM,
and unmarkedFitOccuMulti classes. The character strings supported vary with
the type of model fitted. For unmarkedFitOccu and unmarkedFitOccuMulti
objects, either psi or detect can be supplied to indicate whether the param-
eter is on occupancy or detectability, respectively. For unmarkedFitColExt,
possible values are psi, gamma, epsilon, and detect, for parameters on occu-
For unmarkedFitOccuFP objects, one can specify psi, detect, or fp, for oc-
cupancy, detectability, and probability of assigning false-positives, respectively.
For unmarkedFitOccuRN objects, either lambda or detect can be entered for
abundance and detectability parameters, respectively. For unmarkedFitPCount
and unmarkedFitMPois objects, lambda or detect denote parameters on abun-
dance and detectability, respectively. For unmarkedFitPCO objects, one can
enter lambda, gamma, omega, or detect, to specify parameters on abundance, re-
cruitment, apparent survival, and detectability, respectively. For unmarkedFitDS
objects, only lambda is supported for the moment. For unmarkedFitGDS ob-
jects, lambda and phi denote abundance and availability, respectively. For
unmarkedFitGMM and unmarkedFitGPC objects, lambda, phi, and detect de-
note abundance, availability, and detectability, respectively.
Value
importance returns an object of class importance consisting of the following components:
parm the parameter for which an importance value is required.

w.plus the sum of Akaike weights for the models that include the parameter of interest.
w.minus the sum of Akaike weights for the models that exclude the parameter of interest.
Author(s)
Marc J. Mazerolle
References
Burnham, K. P., and Anderson, D. R. (2002) Model Selection and Multimodel Inference: a practical
importance 113
2248–2255.
84, 2200–2207.
See Also
AICc, aictab, c_hat, confset, evidence, modavg, modavgShrink, modavgPred
Examples
##example on Orthodont data set in nlme
## Not run:
require(nlme)

##random is ~ age | Subject
method = "ML")
Cand.models[[4]] <- lme(distance ~ Sex, data = Orthodont, random = ~ 1,
method = "ML")

importance(cand.set = Cand.models, parm = "age", modnames = Modnames,

second.ord = TRUE, nobs = NULL)
print(importance(cand.set = Cand.models, parm = "age", modnames = Modnames,
second.ord = TRUE, nobs = NULL), digits = 4)
## End(Not run)

## Not run:
require(unmarked)
##single season
data(frogs)
114 importance

obsNum(pferUMF)))

Cand.mods <- list(fm1, fm2, fm3, fm4)
Modnames <- c("fm1", "fm2", "fm3", "fm4")
##compute importance value for 'sitevar1' on occupancy

importance(cand.set = Cand.mods, modnames = Modnames, parm = "sitevar1",
parm.type = "psi")
##compute importance value for 'obsvar1' on detectability
importance(cand.set = Cand.mods, modnames = Modnames, parm = "obsvar1",
parm.type = "detect")
##example with multispecies occupancy modify from ?occuMulti

##Simulate 3 species data
N <- 80
nspecies <- 3
J <- 4
occ_covs <- as.data.frame(matrix(rnorm(N * 10),ncol=10))

names(occ_covs) <- paste('par',1:10,sep='')
det_covs <- list()

for (i in 1:nspecies){
det_covs[[i]] <- matrix(rnorm(N*J),nrow=N)
}
names(det_covs) <- paste('par',1:nspecies,sep='')
##True vals
beta <- c(0.5,0.2,0.4,0.5,-0.1,-0.3,0.2,0.1,-1,0.1)
f1 <- beta[1] + beta[2]*occ_covs$par1
f4 <- beta[7]
f5 <- beta[8]
f6 <- beta[9]
f7 <- beta[10]
f <- cbind(f1,f2,f3,f4,f5,f6,f7)
z <- expand.grid(rep(list(1:0),nspecies))[,nspecies:1]
colnames(z) <- paste('sp',1:nspecies,sep='')
dm <- model.matrix(as.formula(paste0("~.^",nspecies,"-1")),z)
psi <- exp(f

psi <- psi/rowSums(psi)
iron 115
##True state
ztruth <- matrix(NA,nrow=N,ncol=nspecies)
for (i in 1:N){
ztruth[i,] <- as.matrix(z[sample(8,1,prob=psi[i,]),])
}
p_true <- c(0.6,0.7,0.5)
## fake y data
y <- list()
for (i in 1:nspecies){
y[[i]] <- matrix(NA,N,J)
for (j in 1:N){
for (k in 1:J){
y[[i]][j,k] <- rbinom(1,1,ztruth[j,i]*p_true[i])
}
}
}
names(y) <- c('coyote','tiger','bear')
##Create the unmarked data object

data <- unmarkedFrameOccuMulti(y=y,siteCovs=occ_covs,obsCovs=det_covs)
## Formulas for state and detection processes

## Length should match number/order of columns in fDesign
occFormulas <- c('~par1 + par2','~par2','~par3','~1','~1','~1','~1')
occFormulas2 <- c('~par1 + par3','~par1 + par2','~par1 + par2 + par3',
"~ 1", "~1", "~ 1", "~1")
##Length should match number/order of species in data@ylist

detFormulas <- c('~1','~1','~1')
fit <- occuMulti(detFormulas,occFormulas,data)

fit2 <- occuMulti(detFormulas,occFormulas2,data)
##importance
importance(cand.set = list(fit, fit2), parm = "[coyote] par2",
parm.type = "psi")
## End(Not run)
iron Iron Content in Food

116 lizards
Description
This data set, originally from Adish et al. (1999), describes the iron content of food cooked in
different pot types.
Usage
data(iron)
Format
Pot pot type, one of "aluminium", "clay", or "iron".
Food food type, one of "legumes", "meat", or "vegetables".
Iron iron content measured in mg/100 g of food.
Details
Heiberger and Holland (2004, p. 378) use these data as an exercise on two-way ANOVA with
interaction.
Source
Heiberger, R. M., Holland, B. (2004) Statistical Analysis and Data Display: an intermediate course
with examples in S-Plus, R, and SAS. Springer: New York.
Adish, A. A., Esrey, S. A., Gyorkos, T. W., Jean-Baptiste, J., Rojhani, A. (1999) Effect of consump-
tion of food cooked in iron pots on iron status and growth of young children: a randomised trial.
The Lancet 353, 712–716.
Examples
data(iron)
str(iron)
lizards Habitat Preference of Lizards
Description
This data set describes the habitat preference of two species of lizards, Anolis grahami and A.
opalinus, on the island of Jamaica and is originally from Schoener (1970). McCullagh and Nelder
(1989) and Burnham and Anderson (2002) reanalyzed the data. Note that a typo occurs in table
3.11 of Burnham and Anderson (2002).
Usage
data(lizards)
lizards 117
Format
Insolation position of perch, either shaded or sunny.
Diameter diameter of the perch, either < 2 in or >= 2 in.
Height perch height, either < 5 or >= 5.
Time time of day, either morning, midday, or afternoon.
Species species observed, either grahami or opalinus.
Counts number of individuals observed.
Details
Burnham and Anderson (2002, p. 137) use this data set originally from Schoener (1970) to illustrate
model selection for log-linear models.
Source
Hall: New York.
Schoener, T. W. (1970) Nonsynchronous spatial overlap of lizards in patchy habitats. Ecology 51,
408–418.
Examples
data(lizards)
## Not run:
##log-linear model as in Burnham and Anderson 2002, p. 137
##main effects
m1 <- glm(Counts ~ Insolation + Diameter + Height + Time + Species,
family = poisson, data = lizards)
##main effects and all second order interactions = base

m2 <- glm(Counts ~ Insolation + Diameter + Height + Time + Species +
Insolation:Diameter + Insolation:Height + Insolation:Time +
Insolation:Species + Diameter:Height + Diameter:Time +
Diameter:Species + Height:Time + Height:Species +
Time:Species, family = poisson, data = lizards)
##base - DT
Insolation:Species + Diameter:Height + Diameter:Species +
Height:Time + Height:Species + Time:Species,
##base + HDI + HDT + HDS

118 lizards

Time:Species + Height:Diameter:Insolation +
Height:Diameter:Time + Height:Diameter:Species,
##base + HDI + HDS + HIT + HIS + HTS + ITS

Time:Species + Height:Diameter:Insolation +
Height:Diameter:Species + Height:Insolation:Time +
Height:Insolation:Species + Height:Time:Species +
Insolation:Time:Species, family = poisson, data = lizards)
##base + HIT + HIS + HTS + ITS

Time:Species + Height:Insolation:Time +
Height:Insolation:Species + Height:Time:Species +
Insolation:Time:Species, family = poisson, data = lizards)
##base + HIS + HTS + ITS

Time:Species + Height:Insolation:Species +
Height:Time:Species + Insolation:Time:Species,
##base + HIT + HIS + HTS + ITS - DT

Height:Time + Height:Species + Time:Species +
Height:Insolation:Time + Height:Insolation:Species +
Height:Time:Species + Insolation:Time:Species,
##base + HIT + HIS + ITS - DT

Height:Insolation:Time + Height:Insolation:Species +
Insolation:Time:Species,
mb.gof.test 119
##base + HIT + HIS - DT

Height:Insolation:Time + Height:Insolation:Species,
##set up in list
Cands <- list(m1, m2, m3, m4, m5, m6, m7, m8, m9, m10)
Modnames <- paste("m", 1:length(Cands), sep = "")
##model selection
library(AICcmodavg)
aictab(Cands, Modnames)
## End(Not run)
mb.gof.test Compute MacKenzie and Bailey Goodness-of-fit Test for Single Sea-
son, Dynamic, and Royle-Nichols Occupancy Models
Description
These functions compute the MacKenzie and Bailey (2004) goodness-of-fit test for single season
occupancy models based on Pearson’s chi-square and extend it to dynamic (multiple season) and
Royle-Nichols (2003) occupancy models.
Usage
mb.chisq(mod, print.table = TRUE, ...)



mb.chisq(mod, print.table = TRUE, maxK = 50, ...)
mb.gof.test(mod, nsim = 5, plot.hist = TRUE, report = NULL,

parallel = TRUE, ...)

mb.gof.test(mod, nsim = 5, plot.hist = TRUE,
report = NULL, parallel = TRUE, ...)
120 mb.gof.test

report = NULL, parallel = TRUE,

report = NULL, parallel = TRUE, maxK = 50, ...)
Arguments
mod the model for which a goodness-of-fit test is required.
print.table logical. Specifies if the detailed table of observed and expected values is to be
included in the output.
nsim the number of bootstrapped samples.
plot.hist logical. Specifies that a histogram of the bootstrapped test statistic is to be in-
cluded in the output. For dynamic occupancy models, this produces a histogram
of the sum of the season-specific chi-squares for each bootstrap sample.
report If NULL, the test statistic for each iteration is not printed in the terminal. Other-
wise, an integer indicating the number of values of the test statistic that should
be printed on the same line. For example, if report = 3, the values of the test
statistic for three iterations are reported on each line.
parallel logical. If TRUE, requests that parboot use multiple cores to accelerate compu-
tations of the bootstrap.
plot.seasons logical. For dynamic occupancy models, specifies that a histogram of the boot-
strapped test statistic for each primary period (season) is to be included in the
output.
maxK the number of support points used as the summation index in the likelihood of
the Royle-Nichols model (2003). This should match the value used to estimate
the parameters with occuRN.
Details
MacKenzie and Bailey (2004) and MacKenzie et al. (2006) suggest using the Pearson chi-square
to assess the fit of single season occupancy models (MacKenzie et al. 2002). Given low expected
frequencies, the chi-square statistic will deviate from the theoretical distribution and it is recom-
mended to use a parametric bootstrap approach to obtain P-values with the parboot function of the
unmarked package. mb.chisq computes the table of observed and expected values based on the de-
tection histories and single season occupancy model used. mb.gof.test calls internally mb.chisq
and parboot to generate simulated data sets based on the model and compute the MacKenzie and
Bailey test statistic. Missing values are accomodated by creating cohorts for each pattern of missing
values.
It is also possible to obtain an estimate of the overdispersion parameter (c-hat) for the model at hand
by dividing the observed chi-square statistic by the mean of the statistics obtained from simulation.
mb.gof.test 121
This test is extended to dynamic occupancy models of MacKenzie et al. (2003) by using the oc-
cupancy estimates for each season obtained from the model. These estimates are then used to
compute the predicted and observed frequencies separately within each season. The chi-squares are
then summed to be used as the test statistic for the dynamic occupancy model.
Note that values of c-hat > 1 indicate overdispersion (variance > mean), but that values much higher
than 1 (i.e., > 4) probably indicate lack-of-fit. In cases of moderate overdispersion, one usually
multiplies the variance-covariance matrix of the estimates by c-hat. As a result, the SE’s of the
estimates are inflated (c-hat is also known as a variance inflation factor).
In model selection, c-hat should be estimated from the global model and the same value of c-hat
applied to the entire model set. Specifically, a global model is the most complex model from which
all the other models of the set are simpler versions (nested). When no single global model exists in
the set of models considered, such as when sample size does not allow a complex model, one can
estimate c-hat from ’subglobal’ models. Here, ’subglobal’ models denote models from which only
a subset of the models of the candidate set can be derived. In such cases, one can use the smallest
value of c-hat for model selection (Burnham and Anderson 2002).
in package AICcmodavg automatically add 1 when the c.hat argument > 1 and apply the same value
of c-hat for the entire model set. When c-hat > 1, functions compute quasi-likelihood information
criteria (either QAICc or QAIC, depending on the value of the second.ord argument) by scaling the
log-likelihood of the model by c-hat. The value of c-hat can influence the ranking of the models: as
c-hat increases, QAIC or QAICc will favor models with fewer parameters. As an additional check
against this potential problem, one can generate several model selection tables by incrementing
values of c-hat to assess the model selection uncertainty. If ranking changes little up to the c-hat
value observed, one can be confident in making inference.
In cases of underdispersion (c-hat < 1), it is recommended to keep the value of c-hat to 1. However,
note that values of c-hat « 1 can also indicate lack-of-fit and that an alternative model should be
investigated.
Value
mb.chisq returns the following components for single-season and Royle-Nichols occupancy mod-
els:
chisq.table the table of observed and expected values for each detection history and its chi-
square component (if print.table = TRUE). Note that the table only
shows the observed detection histories. Unobserved detection histories are not
shown, but are included in the computation of the test statistic.
chi.square the Pearson chi-square statistic. This test statistic should be compared against a
bootstrap distribution instead of the theoretical chi-square distribution because
low expected frequencies invalidate the chi-square assumption.
model.type the model type, either single-season, royle-nichols, or dynamic.
mb.chisq returns the following additional components for dynamic occupancy models:
tables a list containing the season-specific chi-square tables (if print.table = TRUE).
all.chisq an element containing the season-specific chi-squares.
n.seasons the number of primary periods (seasons).
122 mb.gof.test
mb.gof.test returns the following components for single-season and Royle-Nichols occupancy
models:
chisq.table the table of observed and expected values for each detection history and its chi-
square component.
chi.square the Pearson chi-square statistic.
t.star the bootstrapped chi-square test statistics (i.e., obtained for each of the simulated
data sets).
p.value the P-value assessed from the parametric bootstrap, computed as the proportion
of the simulated test statistics greater than or equal to the observed test statistic.
c.hat.est the estimate of the overdispersion parameter, c-hat, computed as the observed
test statistic divided by the mean of the simulated test statistics.
nsim the number of bootstrap samples. The recommended number of samples varies
with the data set, but should be on the order of 1000 or 5000, and in cases with
a large number of visits, even 10 000 samples, namely to reduce the effect of
unusually small values of the test statistics.
mb.gof.test returns the following additional components for dynamic occupancy models:
chisq.table a list including the table of observed and expected values for each detection
history and its chi-square component for each primary period (season).
chi.square the chi-square test statistic, as the sum of the chi-squares across the primary
periods.
p.value a list of the P-values for each of the primary periods, computed separately as the
proportion of the simulated test statistics greater than or equal to the observed
test statistic.
p.global the P-value of the chi-square test statistic for the dynamic occupancy model.
This P-value is computed as the proportion of the simulated sums of chi-squares
greater than or equal to the observed sum of chi-squares across the primary
periods.
Author(s)
Marc J. Mazerolle
References
2248–2255.
84, 2200–2207.
MacKenzie, D. I., Bailey, L. L. (2004) Assessing the fit of site-occupancy models. Journal of
Agricultural, Biological, and Environmental Statistics 9, 300–318.
mb.gof.test 123
MacKenzie, D. I., Nichols, J. D., Royle, J. A., Pollock, K. H., Bailey, L. L., Hines, J. E. (2006)
Occupancy estimation and modeling: inferring patterns and dynamics of species occurrence. Aca-
demic Press: New York.
Royle, J. A., Nichols, J. D. (2003) Estimating abundance from repeated presence-absence data or
point counts. Ecology 84, 777–790.
See Also
AICc, c_hat, colext, evidence, modavg, importance, modavgPred, Nmix.gof.test, occu, parboot
Examples
## Not run:
require(unmarked)
##single season
data(frogs)

obsNum(pferUMF)))
##run model
##compute observed chi-square

obs <- mb.chisq(fm1)
obs
print(obs, digits.vals = 4)
##compute observed chi-square, assess significance, and estimate c-hat

obs.boot <- mb.gof.test(fm1, nsim = 3)
##note that more bootstrap samples are recommended
##(e.g., 1000, 5000, or 10 000)
obs.boot
print(obs.boot, digits.vals = 4, digits.chisq = 4)
##data with missing values

mat1 <- matrix(c(0, 0, 0), nrow = 120, ncol = 3, byrow = TRUE)
mat2 <- matrix(c(0, 0, 1), nrow = 23, ncol = 3, byrow = TRUE)
mat3 <- matrix(c(1, NA, NA), nrow = 42, ncol = 3, byrow = TRUE)
mat4 <- matrix(c(0, 1, NA), nrow = 33, ncol = 3, byrow = TRUE)
y.mat <- rbind(mat1, mat2, mat3, mat4)
y.sim.data <- unmarkedFrameOccu(y = y.mat)
m1 <- occu(~ 1 ~ 1, data = y.sim.data)
124 min.trap
mb.gof.test(m1, nsim = 3)
##(e.g., 1000, 5000, or 10 000)
## End(Not run)
min.trap Anuran Larvae Counts in Minnow Traps Across Pond Type
Description
This data set consists of counts of anuran larvae as a function of pond type, pond perimeter, and
presence of water scorpions (Ranatra sp.).
Usage
data(min.trap)
Format
Type pond type, denotes the location of ponds in either bog or upland environment
Num_anura number of anuran larvae in minnow traps
Effort number of trap nights (i.e., number of traps x days of trapping) in each pond
Perimeter pond perimeter in meters
Num_ranatra number of water scorpions trapped in minnow traps
log.Perimeter natural log of perimeter
Details
Mazerolle (2006) uses this data set to illustrate model selection for Poisson regression with low
overdispersion.
Source
Examples
data(min.trap)
## maybe str(min.trap) ; plot(min.trap) ...
modavg 125
modavg Compute Model-averaged Parameter Estimate (Multimodel Inference)
Description
This function model-averages the estimate of a parameter of interest among a set of candidate mod-
els, computes the unconditional standard error and unconditional confidence intervals as described
in Buckland et al. (1997) and Burnham and Anderson (2002). This model-averaged estimate is also
referred to as a natural average of the estimate by Burnham and Anderson (2002, p. 152).
Usage
modavg(cand.set, parm, modnames = NULL, second.ord = TRUE, nobs = NULL,
uncond.se = "revised", conf.level = 0.95, exclude = NULL, warn =
TRUE, ...)

modavg(cand.set, parm, modnames = NULL, second.ord =
TRUE, nobs = NULL, uncond.se = "revised", conf.level = 0.95,
exclude = NULL, warn = TRUE, ...)


modavg(cand.set, parm, modnames = NULL,
second.ord = TRUE, nobs = NULL, uncond.se = "revised",
conf.level = 0.95, exclude = NULL, warn = TRUE, ...)


modavg(cand.set, parm, modnames = NULL, second.ord
= TRUE, nobs = NULL, uncond.se = "revised", conf.level = 0.95,


126 modavg


conf.level = 0.95, exclude = NULL, warn = TRUE, c.hat = 1,
gamdisp = NULL, ...)

...)






...)

modavg 127





...)





128 modavg
exclude = NULL, warn = TRUE, c.hat = 1, ...)



modavg(cand.set, parm, modnames =
NULL, second.ord = TRUE, nobs = NULL, uncond.se = "revised",





modavg 129





Arguments
parm the parameter of interest, enclosed between quotes, for which a model-averaged
estimate is required. For a categorical variable, the label of the estimate must be
included as it appears in the output (see ’Details’ below).
models.
rion (i.e., AICc).
130 modavg

value of nobs.
uncond.se either, "old", or "revised", specifying the equation used to compute the un-
conditional standard error of a model-averaged estimate. With uncond.se = "old",
computations are based on equation 4.9 of Burnham and Anderson (2002), which
was the former way to compute unconditional standard errors. With uncond.se =
"revised", equation 6.12 of Burnham and Anderson (2002) is used. Anderson
(2008, p. 111) recommends use of the revised version for the computation of
unconditional standard errors and it is now the default. Note that versions of
package AICcmodavg < 1.04 used the old method to compute unconditional
standard errors.
conf.level the confidence level (1 − α) requested for the computation of unconditional
confidence intervals.
exclude this argument excludes models based on the terms specified for the computation
of a model-averaged estimate of parm. The exclude argument is set to NULL
by default and does not exclude any models other than those without the parm.
When parm is a main effect but is also involved in interactions/polynomial terms
in some models, one should specify the interaction/polynomial terms as a list
to exclude models with these terms from the computation of model-averaged
estimate of the main effect (e.g., exclude = list("sex:mass", "mass2")).
See ’Details’ and ’Examples’ below.
warn logical. If TRUE, modavg performs a check and isssues a warning when the value
in parm occurs more than once in any given model. This is a check for potential
interaction/polynomial terms in the model when such terms are constructed with
the usual operators (e.g., I( ) for polynomial terms, : for interaction terms).
mixture models (Royle 2004, Dail and Madsen 2011). If c.hat > 1, modavg
gamdisp if gamma GLM is used, the dispersion parameter should be specified here to
apply the same value to each model.
unmarkedFitOccuFP, unmarkedFitOccuRN, unmarkedFitMPois, unmarkedFitPCount,
unmarkedFitPCO, unmarkedFitDS, unmarkedFitGDS, unmarkedFitGMM, unmarkedFitGPC,
modavg 131
iota, or detect, to specify parameters on abundance, recruitment, apparent
survival, immigration, and detectability, respectively. For unmarkedFitDS ob-
jects, lambda and detect are supported. For unmarkedFitGDS, lambda, phi,
and detect denote abundance, availability, and detectability, respectively. For
Details
The parameter for which a model-averaged estimate is requested must be specified with the parm
argument and must be identical to its label in the model output (e.g., from summary). For factors,
one must specify the name of the variable and the level of interest. modavg includes checks to find
variations of interaction terms specified in the parm and exclude arguments. However, to avoid
problems, one should specify interaction terms consistently for all models: e.g., either a:b or b:a
for all models, but not a mixture of both.
You must exercise caution when some models include interaction or polynomial terms, because
main effect terms do not have the same interpretation when they also appear in an interaction/polynomial
term in the same model. In such cases, one should exclude models containing interaction terms
where the main effect is involved with the exclude argument of modavg. Note that modavg checks
for potential cases of multiple instances of a variable appearing more than once in a given model
(presumably in an interaction) and issues a warning. To correctly compute the model-averaged es-
timate of a main effect involved in interaction/polynomial terms, specify the interaction terms(s)
that should not appear in the same model with the exclude argument. This will effectively exclude
models from the computation of the model-averaged estimate.
When warn = TRUE, modavg looks for matches among the labels of the estimates with identical.
It then compares the results to partial matches with regexpr, and issues a warning whenever they
are different. As a result, modavg may issue a warning when some variables or levels of categorical
variables have nested names (e.g., treat, treat10; L, TL). When this warning is only due to the
presence of similarly named variables in the models (and NOT due to interaction terms), you can
suppress this warning by setting warn = FALSE.
The model-averaging estimator implemented in modavg is known to be biased away from 0 when
there is substantial model selection uncertainty (Cade 2015). In such instances, it is recommended
to use the model-averaging shrinkage estimator (i.e., modavgShrink) for inference on beta esti-
mates or to focus on model-averaged effect sizes (modavgEffect) and model-averaged predictions
(modavgPred).
modavg is implemented for a list containing objects of aov, betareg, clm, clmm, clogit, coxme,
coxph, glm, glmmTMB, gls, hurdle, lm, lme, lmekin, maxlikeFit, mer, glmerMod, lmerMod,
lmerModLmerTest, multinom, polr, rlm, survreg, vglm, zeroinfl classes as well as various
models of unmarkedFit classes.
132 modavg
Value
modavg creates an object of class modavg with the following components:
Parameter the parameter for which a model-averaged estimate was obtained.
Mod.avg.table the reduced model selection table based on models including the parameter of
interest.
Mod.avg.beta the model-averaged estimate based on all models including the parameter of
interest (see ’Details’ above regarding the exclusion of models where parameter
of interest is involved in an interaction).
Uncond.SE the unconditional standard error for the model-averaged estimate (as opposed to
the conditional SE based on a single model).
Conf.level the confidence level used to compute the confidence interval.
Lower.CL the lower confidence limit.
Upper.CL the upper confidence limit.
Author(s)
Marc J. Mazerolle
References
New York.
Buckland, S. T., Burnham, K. P., Augustin, N. H. (1997) Model selection: an integral part of
inference. Biometrics 53, 603–618.
Cade, B. S. (2015) Model averaging and muddled multimodel inferences. Ecology 96, 2370–2382.
2248–2255.
84, 2200–2207.
modavg 133
See Also
AICc, aictab, c_hat, confset, evidence, importance, modavgCustom, modavgEffect, modavgShrink,
modavgPred
Examples
##anuran larvae example modified from Mazerolle (2006)
##these are different models than in the paper
data(min.trap)

Cand.mod <- list( )
##global model
Cand.mod[[1]] <- glm(Num_anura ~ Type + log.Perimeter +
Type:log.Perimeter + Num_ranatra,
data = min.trap)
##interactive model
Cand.mod[[2]] <- glm(Num_anura ~ Type + log.Perimeter +
Type:log.Perimeter, family = poisson,
##additive model
##Predator model

c_hat(Cand.mod[[1]]) #uses Pearson's chi-square/df
##to 1

##model selection
aictab(Cand.mod, Modnames)
##compute model-averaged estimates for parameters appearing in top

##models
modavg(parm = "Num_ranatra", cand.set = Cand.mod, modnames = Modnames)
print(modavg(parm = "Num_ranatra", cand.set = Cand.mod,
modnames = Modnames), digits = 4)
##model-averaging a variable involved in an interaction

##the following produces an error - because the variable is involved
134 modavg
##in an interaction in some candidate models

## Not run: modavg(parm = "TypeBOG", cand.set = Cand.mod,
modnames = Modnames)
## End(Not run)
##exclude models where the variable is involved in an interaction

##to get model-averaged estimate of main effect
modavg(parm = "TypeBOG", cand.set = Cand.mod, modnames = Modnames,
exclude = list("Type:log.Perimeter"))
##to get model-averaged estimate of interaction

modavg(parm = "TypeBOG:log.Perimeter", cand.set = Cand.mod,
##beware of variables that have similar names

set.seed(seed = 4)
resp <- rnorm(n = 40, mean = 3, sd = 1)
size <- rep(c("small", "medsmall", "high", "medhigh"), times = 10)
set.seed(seed = 4)
mass <- rnorm(n = 40, mean = 2, sd = 0.1)
mass2 <- mass^2
age <- rpois(n = 40, lambda = 3.2)
agecorr <- rpois(n = 40, lambda = 2)
sizecat <- rep(c("a", "ab"), times = 20)
data1 <- data.frame(resp = resp, size = size, sizecat = sizecat,
mass = mass, mass2 = mass2, age = age,
agecorr = agecorr)
##set up models in list

Cand <- list( )
Cand[[1]] <- lm(resp ~ size + agecorr, data = data1)
Cand[[2]] <- lm(resp ~ size + mass + agecorr, data = data1)
Cand[[3]] <- lm(resp ~ age + mass, data = data1)
Cand[[4]] <- lm(resp ~ age + mass + mass2, data = data1)
Cand[[5]] <- lm(resp ~ mass + mass2 + size, data = data1)
Cand[[6]] <- lm(resp ~ mass + mass2 + sizecat, data = data1)
Cand[[7]] <- lm(resp ~ sizecat, data = data1)
Cand[[8]] <- lm(resp ~ sizecat + mass + sizecat:mass, data = data1)
Cand[[9]] <- lm(resp ~ agecorr + sizecat + mass + sizecat:mass,
data = data1)
##create vector of model names

Modnames <- paste("mod", 1:length(Cand), sep = "")
aictab(cand.set = Cand, modnames = Modnames, sort = TRUE) #correct
##as expected, issues warning as mass occurs sometimes with "mass2" or

##"sizecatab:mass" in some of the models
## Not run: modavg(cand.set = Cand, parm = "mass", modnames = Modnames)
modavg 135
##no warning issued, because "age" and "agecorr" never appear in same model
modavg(cand.set = Cand, parm = "age", modnames = Modnames)
##as expected, issues warning because warn=FALSE, but it is a very bad

##idea in this example since "mass" occurs with "mass2" and "sizecat:mass"
##in some of the models - results are INCORRECT
## Not run: modavg(cand.set = Cand, parm = "mass", modnames = Modnames,
warn = FALSE)
## End(Not run)
##correctly excludes models with quadratic term and interaction term

##results are CORRECT
modavg(cand.set = Cand, parm = "mass", modnames = Modnames,
exclude = list("mass2", "sizecat:mass"))
##correctly computes model-averaged estimate because no other parameter

##occurs simultaneously in any of the models
modavg(cand.set = Cand, parm = "sizesmall", modnames = Modnames) #correct
##as expected, issues a warning because "sizecatab" occurs sometimes in

##an interaction in some models
## Not run: modavg(cand.set = Cand, parm = "sizecatab",
## End(Not run)
##exclude models with "sizecat:mass" interaction - results are CORRECT

modavg(cand.set = Cand, parm = "sizecatab", modnames = Modnames,
exclude = list("sizecat:mass"))
##example with multiple-season occupancy model modified from ?colext

##this is a bit longer
## Not run:
require(unmarked)
data(frogs)
umf <- formatMult(masspcru)
obsCovs(umf) <- scale(obsCovs(umf))
siteCovs(umf) <- rnorm(numSites(umf))
yearlySiteCovs(umf) <- data.frame(year = factor(rep(1:7,
numSites(umf))))
##set up model with constant transition rates

fm <- colext(psiformula = ~ 1, gammaformula = ~ 1, epsilonformula = ~ 1,
pformula = ~ JulianDate + I(JulianDate^2), data = umf,
control = list(trace=1, maxit=1e4))
##model with with year-dependent transition rates

fm.yearly <- colext(psiformula = ~ 1, gammaformula = ~ year,
epsilonformula = ~ year,
pformula = ~ JulianDate + I(JulianDate^2),
data = umf)
136 modavg.utility
##store in list and assign model names

Cand.mods <- list(fm, fm.yearly)
Modnames <- c("psi1(.)gam(.)eps(.)p(Date + Date2)",
"psi1(.)gam(Year)eps(Year)p(Date + Date2)")
##compute model-averaged estimate of occupancy in the first year

modavg(cand.set = Cand.mods, modnames = Modnames, parm = "(Intercept)",
parm.type = "psi")
##compute model-averaged estimate of Julian Day squared on detectability

modavg(cand.set = Cand.mods, modnames = Modnames,
parm = "I(JulianDate^2)", parm.type = "detect")
## End(Not run)
##example of model-averaged estimate of area from distance model

## Not run:
data(linetran) #example modified from ?distsamp

dist.breaks = c(0, 5, 10, 15, 20),
tlength = linetran$Length * 1000, survey = "line", unitsIn = "m")
})

## Halfnormal. Covariates affecting both density and detection.

fm2 <- distsamp(~ area + habitat ~ area + habitat, ltUMF)
## Hazard function. Covariates affecting both density and detection.

fm3 <- distsamp(~ habitat ~ area + habitat, ltUMF, keyfun="hazard")

Modnames <- paste("mod", 1:length(Cands), sep = "")
##model-average estimate of area on abundance

modavg(cand.set = Cands, modnames = Modnames, parm = "area", parm.type = "lambda")
## End(Not run)
modavg.utility Various Utility Functions

modavg.utility 137
Description
reverse.parm and reverse.exclude reverse the order of variables in an interaction term.
formatCands creates new classes for lists containing candidate models.
formulaShort prints a succinct formula from an unmarkedFit object.
Usage
reverse.parm(parm)
reverse.exclude(exclude)
formatCands(cand.set)
formulaShort(mod, unmarked.type = NULL)
Arguments
parm a parameter to be model-averaged, enclosed between quotes, as it appears in the
output of some models.
exclude a list of interaction or polynomial terms appearing in some models, as they
would appear in the call to the model function (i.e., A*B, A:B). Models contain-
ing elements from the list will be excluded to obtain a model-averaged estimate.
mod an object storing the result of an unmarkedFit.
unmarked.type a character string specifying the type of parameter in an unmarkedFit for which
a formula is requested. This argument uses the character string for each param-
eter group as defined by unmarked.
Details
These utility functions are used internally by aictab, modavg, and other related functions.
reverse.parm and reverse.exclude enable the user to specify differently interaction terms (e.g.,
A:B, B:A) across models for model averaging. These functions have been added to avoid problems
when users are not consistent in the specification of interaction terms across models.
formatCands creates new classes for the list of candidate models based on the contents of the list.
These new classes are used for method dispatch.
formulaShort is used by anovaOD.
Value
reverse.parm returns all possible combinations of an interaction term to identify models that in-
clude the parm of interest and find the corresponding estimate and standard error in the model
object.
reverse.exclude returns a list of all possible combinations of exclude to identify models that
should be excluded when computing a model-averaged estimate.
formatCands adds a new class to the list of candidate models based on the classes of the models.
formulaShort creates a character string for the formula related to a given parameter type from an
unmarkedFit object.
138 modavg.utility
Author(s)
Marc J. Mazerolle
See Also
aictab, anovaOD, modavg, modavgShrink, modavgPred
Examples
##a main effect
reverse.parm(parm = "Ageyoung") #does not return anything
##an interaction term as it might appear in the output

reverse.parm(parm = "Ageyoung:time") #returns the reverse
##exclude two interaction terms

reverse.exclude(exclude = list("Age*time", "A:B"))
##returns all combinations
reverse.exclude(exclude = list("Age:time", "A*B"))
##returns all combinations

data(dry.frog)

data = dry.frog)
formatCands(Cand.models)
## Not run:
require(unmarked)
data(bullfrog)
bfrog <- unmarkedFrameOccu(y = bullfrog[, c("V1", "V2", "V3", "V4")],
siteCovs = bullfrog[, 1:2])
fm1 <- occu(~ 1 ~ Reed.presence, data = bfrog)
formulaShort(fm1, unmarked.type = "state")
formulaShort(fm1, unmarked.type = "det")
## End(Not run)
modavgCustom 139
modavgCustom Compute Model-averaged Parameter Estimate from User-supplied In-

put Based on (Q)AIC(c)
Description
This function model-averages the estimate of a parameter of interest among a set of candidate
models, and computes the unconditional standard error and unconditional confidence intervals as
described in Buckland et al. (1997) and Burnham and Anderson (2002).
Usage
modavgCustom(logL, K, modnames = NULL, estimate, se, second.ord = TRUE,
nobs = NULL, uncond.se = "revised", conf.level = 0.95,
c.hat = 1, useBIC = FALSE)
Arguments
models.
estimate a vector of estimates for each of the models in the candidate model set. Estimates
can be either beta estimates for a parameter of interest or a single prediction from
each model.
se a vector of standard errors for each of the estimates appearing in the estimate
vector.
rion (i.e., AICc). This argument is ignored if useBIC = TRUE.
nobs the sample size required to compute the AICc, QAICc, BIC, or QBIC.
unconditional standard errors and it is now the default.
140 modavgCustom
obtained from c_hat. Note that values of c.hat different from 1 are only ap-
propriate for binomial GLM’s with trials > 1 (i.e., success/trial or cbind(success,
failure) syntax), with Poisson GLM’s, single-season and dynamic occupancy
models (MacKenzie et al. 2002, 2003), N-mixture models (Royle 2004, Dail and
Madsen 2011), or capture-mark-recapture models (e.g., Lebreton et al. 1992).
If c.hat > 1, modavgCustom will return the quasi-likelihood analogue of the in-
formation criteria requested and multiply the variance-covariance matrix of the
estimates by this value (i.e., SE’s are multiplied by sqrt(c.hat)).
useBIC logical. If TRUE, the function returns the Bayesian information criterion (BIC)
when c.hat = 1 or the quasi-likelihood BIC (QBIC) when c.hat > 1.
Details
modavgCustom computes a model-averaged estimate from the vector of parameter estimates speci-
fied in estimate. Estimates and their associated standard errors must be specified in the same order
as the log-likelihood, number of estimated parameters, and model names. Estimates provided may
be for a parameter of interest (i.e., beta estimates) or predictions from each model. This function
is most useful when model input is imported into R from other software (e.g., Program MARK,
PRESENCE) or for model classes that are not yet supported by the other model averaging functions
such as modavg or modavgPred.
Value
modavgCustom creates an object of class modavgCustom with the following components:
Mod.avg.table the model selection table

Mod.avg.est the model-averaged estimate
Uncond.SE the unconditional standard error for the model-averaged estimate
Conf.level the confidence level used to compute the confidence interval
Lower.CL the lower confidence limit
Upper.CL the upper confidence limit
Author(s)
Marc J. Mazerolle
References
New York.
modavgCustom 141
2248–2255.
84, 2200–2207.
See Also
AICcCustom, aictabCustom, bictabCustom, modavg, modavgIC, modavgShrink, modavgPred
Examples
## Not run:
##model averaging parameter estimate (natural average)
LL <- c(-38.8876, -35.1783, -64.8970)

Ks <- c(7, 9, 4)

##vector of beta estimates for a parameter of interest

model.ests <- c(0.0478, 0.0480, 0.0478)
##vector of SE's of beta estimates for a parameter of interest

model.se.ests <- c(0.0028, 0.0028, 0.0034)
##compute model-averaged estimate and unconditional SE based on AICc

estimate = model.ests, se = model.se.ests, nobs = 121)
##compute model-averaged estimate and unconditional SE based on BIC
estimate = model.ests, se = model.se.ests, nobs = 121,
useBIC = TRUE)
##model-averaging with shrinkage based on AICc

data(min.trap)
Cand.mod <- list( )
##global model
Cand.mod[[1]] <- glm(Num_anura ~ Type + log.Perimeter,
142 modavgEffect
data = min.trap)
family = poisson, offset = log(Effort), data = min.trap)
Model.names <- c("Type + log.Perimeter", "Type + Num_ranatra",
"log.Perimeter + Num_ranatra")
##model-averaged estimate with shrinkage (glm model type is already supported)
modavgShrink(cand.set = Cand.mod, modnames = Model.names,
parm = "log.Perimeter")
##equivalent manual version of model-averaging with shrinkage

##this is especially useful when model classes are not supported
##extract vector of LL
LLs <- sapply(Cand.mod, FUN = function(i) logLik(i)[1])
##extract vector of K
Ks <- sapply(Cand.mod, FUN = function(i) attr(logLik(i), "df"))
##extract betas
betas <- sapply(Cand.mod, FUN = function(i) coef(i)["log.Perimeter"])
##second model does not include log.Perimeter
betas[2] <- 0
##extract SE's
ses <- sapply(Cand.mod, FUN = function(i) sqrt(diag(vcov(i))["log.Perimeter"]))
ses[2] <- 0
##model-averaging with shrinkage based on AICc
modavgCustom(logL = LLs, K = Ks, modnames = Model.names,
nobs = nrow(min.trap), estimate = betas, se = ses)
##model-averaging with shrinkage based on BIC
modavgCustom(logL = LLs, K = Ks, modnames = Model.names,
nobs = nrow(min.trap), estimate = betas, se = ses,
useBIC = TRUE)
## End(Not run)
modavgEffect Compute Model-averaged Effect Sizes (Multimodel Inference on

Group Differences)
Description
This function model-averages the effect size between two groups defined by a categorical variable
based on the entire model set and computes the unconditional standard error and unconditional
confidence intervals as described in Buckland et al. (1997) and Burnham and Anderson (2002).
This can be particularly useful when dealing with data from an experiment (e.g., ANOVA) and when
the focus is to determine the effect of a given factor. This is an information-theoretic alternative to
multiple comparisons (e.g., Burnham et al. 2011).
Usage
modavgEffect(cand.set, modnames = NULL, newdata, second.ord = TRUE,
modavgEffect 143

...)

modavgEffect(cand.set, modnames = NULL, newdata,
conf.level = 0.95, ...)

conf.level = 0.95, type = "response", c.hat = 1, gamdisp = NULL,
...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, type = "response", ...)

modavgEffect(cand.set, modnames = NULL,
newdata, second.ord = TRUE, nobs = NULL, uncond.se = "revised",

conf.level = 0.95, ...)

conf.level = 0.95, ...)
144 modavgEffect


conf.level = 0.95, ...)


conf.level = 0.95, type = "response", c.hat = 1,

modavgEffect(cand.set, modnames =
NULL, newdata, second.ord = TRUE, nobs = NULL, uncond.se =
"revised", conf.level = 0.95, type = "response",
c.hat = 1, parm.type = NULL, ...)




modavgEffect 145






Arguments
models.
newdata a data frame with two rows and where the columns correspond to the explana-
tory variables specified in the candidate models. Note that this data set must
have the same structure as that of the original data frame for which we want to
make predictions, specifically, the same variable type and names that appear in
the original data set. Each row of the data set defines one of the two groups com-
pared. The first row in newdata defines the first group, whereas the second row
146 modavgEffect
defines the second group. The effect size is computed as the prediction in the
first row minus the prediction in the second row (first row - second row). Only
the column relating to the grouping variable can change value and all others
must be held constant for the comparison (see ’Details’).
rion (i.e., AICc).
nobs this argument allows the specification of a numeric value other than total sample
size to compute the AICc (i.e., nobs defaults to total number of observations).
This is relevant only for mixed models or various models of unmarkedFit classes
where sample size is not straightforward. In such cases, one might use total num-
ber of observations or number of independent clusters (e.g., sites) as the value
of nobs.
standard errors.
conf.level the confidence level (1 − α) requested for the computation of unconditional con-
fidence intervals. To obtain confidence intervals corrected for multiple compar-
isons between pairs of treatments, it is possible to adjust the α level according
to various strategies such as the Bonferroni correction (Dunn 1961).
type the scale of prediction requested, one of "response" or "link" (only relevant
for glm, mer, and unmarkedFit classes). Note that the value "terms" is not
defined for modavgEffect).
ure) syntax), with Poisson GLM’s, single-season and dynamic occupancy mod-
els (MacKenzie et al. 2002, 2003), or N-mixture models (Royle 2004, Dail and
Madsen 2011). If c.hat > 1, modavgEffect will return the quasi-likelihood
analogue of the information criteria requested and multiply the variance-covariance
matrix of the estimates by this value (i.e., SE’s are multiplied by sqrt(c.hat)).
This option is not supported for generalized linear mixed models of the mer
class.
unmarkedFitPCO, unmarkedFitDS, unmarkedFitGDS, unmarkedFitGMM, and unmarkedFitGPC
classes. The character strings supported vary with the type of model fitted.
For unmarkedFitOccu objects, either psi or detect can be supplied to indi-
cate whether the parameter is on occupancy or detectability, respectively. For
modavgEffect 147
unmarkedFitColExt, possible values are psi, gamma, epsilon, and detect,

for parameters on occupancy in the inital year, colonization, extinction, and
detectability, respectively. For unmarkedFitOccuFP objects, one can specify
psi, detect, falsepos, or certain, for occupancy, detectability, probability
of assigning false-positives, and probability detections are certain, respectively.
enter lambda, gamma, omega, iota, or detect, to specify parameters on abun-
dance, recruitment, apparent survival, immigration, and detectability, respec-
tively. For unmarkedFitDS objects, lambda and detect are supported. For
unmarkedFitGDS, lambda, phi, and detect denote abundance, availability, and
detectability, respectively. For unmarkedFitGMM and unmarkedFitGPC objects,
lambda, phi, and detect denote abundance, availability, and detectability, re-
spectively.
Details
The strategy used here to compute effect sizes is to work from the newdata object to create two
predictions from a given model and compute the differences and standard errors between both val-
ues. This step is executed for each model in the candidate model set, to obtain a model-averaged
estimate of the effect size and unconditional standard error. As a result, the newdata argument is
restricted to two rows, each for a given prediction. To specify each group, the values entered in the
column for each explanatory variable can be identical, except for the grouping variable. In such a
case, the function will identify the variable and the assign group names based on the values of the
variable. If more than a single variable has different values in its respective column, the function
will print generic names in the output to identify the two groups. A sensible choice of value for the
explanatory variables to be held constant is the average of the variable.
Model-averaging effect sizes is most useful in true experiments (e.g., ANOVA-type designs), where
one wants to obtain the best estimate of effect size given the support of each candidate model. This
can be considered as a information-theoretic analog of traditional multiple comparisons, except
that the information contained in the entire model set is used instead of being restricted to a single
model. See ’Examples’ below for applications.
modavgEffect calls the appropriate method depending on the class of objects in the list. The cur-
rent classes supported include aov, glm, gls, lm, lme, mer, glmerMod, lmerMod, lmerModLmerTest,
rlm, survreg, as well as unmarkedFitOccu, unmarkedFitColExt, unmarkedFitOccuFP, unmarkedFitOccuRN,
unmarkedFitPCount, unmarkedFitPCO, unmarkedFitDS, unmarkedFitGDS, unmarkedFitMPois,
unmarkedFitGMM, and unmarkedFitGPC classes.
Value
The result is an object of class modavgEffect with the following components:
Group.variable the grouping variable defining the two groups compared.

Group1 the first group considered in the comparison.
Group2 the second group considered in the comparison.
148 modavgEffect
Type the scale on which the model-averaged effect size was computed (e.g., response
or link).
Mod.avg.table the full model selection table including the entire set of candidate models.
Mod.avg.eff the model-averaged effect size based on the entire candidate model set.
Uncond.SE the unconditional standard error for the model-averaged effect size.
Author(s)
Marc J. Mazerolle
References
New York.
Dunn, O. J. (1961) Multiple comparisons among means. Journal of the American Statistical Asso-
ciation 56, 52–64.
2248–2255.
84, 2200–2207.
See Also
AICc, aictab, c_hat, confset, evidence, importance, modavgShrink, modavgPred
modavgEffect 149
Examples
##heights (cm) of plants grown under two fertilizers, Ex. 9.5 from
##Zar (1984): Biostatistical Analysis. Prentice Hall: New Jersey.
heights <- data.frame(Height = c(48.2, 54.6, 58.3, 47.8, 51.4, 52.0,
55.2, 49.1, 49.9, 52.6, 52.3, 57.4, 55.6, 53.2,
61.3, 58.0, 59.8, 54.8),
Fertilizer = c(rep("old", 10), rep("new", 8)))
##run linear model hypothesizing an effect of fertilizer

m1 <- lm(Height ~ Fertilizer, data = heights)
##run null model (no effect of fertilizer)

m0 <- lm(Height ~ 1, data = heights)
##assemble models in list

Cands <- list(m1, m0)
Modnames <- c("Fert", "null")
##compute model selection table to compare

##both hypotheses
aictab(cand.set = Cands, modnames = Modnames)
##note that model with fertilizer effect is much better supported
##than the null
##compute model-averaged effect sizes: one model hypothesizes a

##difference of 0, whereas the other assumes a difference
##prepare newdata object from which differences between groups

##will be computed
##the first row of the newdata data.frame relates to the first group,
##whereas the second row corresponds to the second group
pred.data <- data.frame(Fertilizer = c("new", "old"))
##compute best estimate of effect size accounting for model selection

##uncertainty
modavgEffect(cand.set = Cands, modnames = Modnames,
newdata = pred.data)
##classical one-way ANOVA type-design

## Not run:
##generate data for two groups and control
set.seed(seed = 15)
y <- round(c(rnorm(n = 15, mean = 10, sd = 5),
rnorm(n = 15, mean = 15, sd = 5),
rnorm(n = 15, mean = 12, sd = 5)), digits = 2)
##groups
group <- c(rep("cont", 15), rep("trt1", 15), rep("trt2", 15))
##combine in data set

aov.data <- data.frame(Y = y, Group = group)
rm(y, group)
150 modavgEffect
##run model with group effect

lm.eff <- lm(Y ~ Group, data = aov.data)
##null model
lm.0 <- lm(Y ~ 1, data = aov.data)
##compare both models

Cands <- list(lm.eff, lm.0)
Mods <- c("group effect", "no group effect")
aictab(cand.set = Cands, modnames = Mods)
##model with group effect has most of the weight
##compute model-averaged effect sizes

##trt1 - control
newdata = data.frame(Group = c("trt1", "cont")))
##trt1 differs from cont
##trt2 - control
newdata = data.frame(Group = c("trt2", "cont")))
##trt2 does not differ from cont
## End(Not run)
##two-way ANOVA type design, Ex. 13.1 (Zar 1984) of plasma calcium
##concentration (mg/100 ml) in birds as a function of sex and hormone
##treatment
## Not run:
birds <- data.frame(Ca = c(16.87, 16.18, 17.12, 16.83, 17.19, 15.86,
14.92, 15.63, 15.24, 14.8, 19.07, 18.77, 17.63,
16.99, 18.04, 17.2, 17.64, 17.89, 16.78, 16.92,
32.45, 28.71, 34.65, 28.79, 24.46, 30.54, 32.41,
28.97, 28.46, 29.65),
Sex = c("M", "M", "M", "M", "M", "F", "F", "F", "F",
"F", "M", "M", "M", "M", "M", "F", "F", "F", "F",
"F", "M", "M", "M", "M", "M", "F", "F", "F", "F",
"F"),
Hormone = as.factor(c(1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3,
3, 3, 3, 3, 3)))
##candidate models
##interactive effects
m.inter <- lm(Ca ~ Sex + Hormone + Sex:Hormone, data = birds)
##additive effects
m.add <- lm(Ca ~ Sex + Hormone, data = birds)
##Sex only
m.sex <- lm(Ca ~ Sex, data = birds)
modavgEffect 151
##Hormone only
m.horm <- lm(Ca ~ Hormone, data = birds)
##null
m.0 <- lm(Ca ~ 1, data = birds)
##model selection
Cands <- list(m.inter, m.add, m.sex, m.horm, m.0)
Mods <- c("interaction", "additive", "sex only", "horm only", "null")
aictab(Cands, Mods)
##there is some support for a hormone only treatment, but also for
##additive effects
##compute model-averaged effects of sex, and set the other variable

##to a constant value
##M - F
sex.data <- data.frame(Sex = c("M", "F"), Hormone = c("1", "1"))
modavgEffect(Cands, Mods, newdata = sex.data)
##no support for a sex main effect
##hormone 1 - 3, but set Sex to a constant value

horm1.data <- data.frame(Sex = c("M", "M"), Hormone = c("1", "3"))
modavgEffect(Cands, Mods, newdata = horm1.data)
##hormone 2 - 3, but set Sex to a constant value

horm2.data <- data.frame(Sex = c("M", "M"), Hormone = c("2", "3"))
modavgEffect(Cands, Mods, newdata = horm2.data)
## End(Not run)
##Poisson regression with anuran larvae example from Mazerolle (2006)

## Not run:
data(min.trap)

Cand.mod <- list( )
##global model
Cand.mod[[1]] <- glm(Num_anura ~ Type + log.Perimeter,
data = min.trap)

vif.hat <- c_hat(Cand.mod[[1]]) #uses Pearson's chi-square/df
152 modavgEffect

Modnames <- c("type + logperim", "type", "logperim", "intercept only")
##compute model-averaged estimate of difference between abundance at bog

##pond and upland pond
##create newdata object to make predictions
pred.data <- data.frame(Type = c("BOG", "UPLAND"),
log.Perimeter = mean(min.trap$log.Perimeter),
Effort = mean(min.trap$Effort))
modavgEffect(Cand.mod, Modnames, newdata = pred.data, c.hat = vif.hat,
type = "response")
##little suport for a pond type effect
## End(Not run)
##mixed linear model example from ?nlme

## Not run:
library(nlme)
Cand.models[[1]] <- lme(distance ~ age, data = Orthodont, method="ML")
random = ~ 1, method="ML")
Cand.models[[3]] <-lme(distance ~ 1, data = Orthodont, random = ~ 1,
method="ML")
Cand.models[[4]] <-lme(distance ~ Sex, data = Orthodont, random = ~ 1,
method="ML")
Modnames <- c("age", "age + sex", "null", "sex")
data.other <- data.frame(age = mean(Orthodont$age),

Sex = factor(c("Male", "Female")))
modavgEffect(cand.set = Cand.models, modnames = Modnames,
newdata = data.other, conf.level = 0.95, second.ord = TRUE,
nobs = NULL, uncond.se = "revised")
## End(Not run)
##site occupancy analysis example

## Not run:
library(unmarked)
##single season model
data(frogs)
##create a bogus site group
site.group <- c(rep(1, times = nrow(pfer.bin)/2), rep(0, nrow(pfer.bin)/2))

siteCovs(pferUMF) <- data.frame(site.group, sitevar1 =
rnorm(numSites(pferUMF)),
modavgEffect 153

obsCovs(pferUMF) <- data.frame(obsvar1 =
rnorm(numSites(pferUMF) * obsNum(pferUMF)))
fm1 <- occu(~ obsvar1 ~ site.group, pferUMF)

fm2 <- occu(~ obsvar1 ~ 1, pferUMF)
Cand.mods <- list(fm1, fm2)

Modnames <- c("fm1", "fm2")

aictab(cand.set = Cand.mods, modnames = Modnames, second.ord = TRUE)
##model-averaged effect sizes comparing site.group 1 - site.group 0

newer.dat <- data.frame(site.group = c(0, 1))
modavgEffect(cand.set = Cand.mods, modnames = Modnames, type = "response",

second.ord = TRUE, newdata = newer.dat, parm.type = "psi")
##no support for an effect of site group
## End(Not run)
##single season N-mixture models

## Not run:
data(mallard)
##this variable was created to illustrate the use of modavgEffect
##with detection variables
mallard.site$site.group <- c(rep(1, 119), rep(0, 120))
siteCovs(mallardUMF)
tmp.covs <- obsCovs(mallardUMF)
obsCovs(mallardUMF)$date2 <- tmp.covs$date^2
(fm.mall <- pcount(~ site.group ~ length + elev + forest, mallardUMF, K=30))
(fm.mallb <- pcount(~ 1 ~ length + elev + forest, mallardUMF, K=30))
Cands <- list(fm.mall, fm.mallb)

Modnames <- c("one", "null")
##model averaged effect size of site.group 1 - site.group 0 on response

##scale (point estimate)
modavgEffect(Cands, Modnames, newdata = data.frame(site.group = c(0, 1)),
parm.type = "detect", type = "response")
##model averaged effect size of site.group 1 - site.group 0 on link

##scale (here, logit link)
modavgEffect(Cands, Modnames, newdata = data.frame(site.group = c(0, 1)),
parm.type = "detect", type = "link")
154 modavgIC
## End(Not run)
modavgIC Compute Model-averaged Parameter Estimate from User-supplied In-

formation Criterion
Description
This function model-averages the estimate of a parameter of interest among a set of candidate
models, and computes the unconditional standard error and unconditional confidence intervals as
described in Buckland et al. (1997) and Burnham and Anderson (2002). Computations are based
on the values of the information criterion supplied manually by the user.
Usage
modavgIC(ic, K, modnames = NULL, estimate, se, uncond.se = "revised",
conf.level = 0.95, ic.name = NULL)
Arguments
ic a vector of information criterion values for each model in the candidate model
set.
modnames a character vector of model names to identify each model in the model selection
table. If NULL, generic names (e.g., Mod1, Mod2) are supplied in the table in the
same order as the information criterion values.
estimate a vector of estimates for each of the models in the candidate model set. Estimates
can be either beta estimates for a parameter of interest or a single prediction from
each model.
se a vector of standard errors for each of the estimates appearing in the estimate
vector.
unconditional standard errors and it is now the default.
ic.name a character string denoting the name of the information criterion input by the
user. This character string will appear in certain column labels of the model
selection table.
modavgIC 155
Details
modavgIC computes a model-averaged estimate from the vector of parameter estimates specified in
estimate. Estimates and their associated standard errors must be specified in the same order as
the values of the information criterion, the number of estimated parameters, and the model names.
Estimates provided may be for a parameter of interest (i.e., beta estimates) or predictions from each
model. This function is most useful for information criterion other than AIC, AICc, QAIC, and
QAICc (e.g., WAIC: Watanabe 2010) or for classes not supported by modavg, modavgCustom, or
modavgPred.
Value
modavgIC creates an object of class modavgIC with the following components:
Mod.avg.table the model selection table.

Mod.avg.est the model-averaged estimate.
Uncond.SE the unconditional standard error for the model-averaged estimate.
Author(s)
Marc J. Mazerolle
References
New York.
Watanabe, S. (2010) Asymptotic equivalence of Bayes cross validation and widely applicable infor-
mation criterion in singular learning theory. Journal of Machine Learning Research 11, 3571–3594.
See Also
aictabCustom, ictab, modavg, modavgCustom, modavgShrink, modavgPred
Examples
## Not run:
##model averaging parameter estimate based on WAIC
156 modavgPred
##WAIC values
waic <- c(105.74, 107.36, 108.24, 100.57)
effK <- c(7.45, 5.61, 6.14, 6.05)
##vector of predictions
Preds <- c(0.106, 0.137, 0.067, 0.050)
##vector of SE's for prediction
Ses <- c(0.128, 0.159, 0.054, 0.039)
##compute model-averaged estimate and unconditional SE based on WAIC

modavgIC(ic = waic, K = effK, modnames = Modnames,
ic.name = "WAIC")
## End(Not run)
modavgPred Compute Model-averaged Predictions
Description
This function computes the model-averaged predictions, unconditional standard errors, and confi-
dence intervals based on the entire candidate model set. The function is currently implemented for
glm, gls, lm, lme, mer, merMod, lmerModLmerTest, negbin, rlm, survreg object classes that are
stored in a list as well as various models of unmarkedFit classes.
Usage
modavgPred(cand.set, modnames = NULL, newdata, second.ord = TRUE,
nobs = NULL, uncond.se = "revised", conf.level = 0.95, ...)

modavgPred(cand.set, modnames = NULL, newdata,
conf.level = 0.95, ...)

gamdisp = NULL, ...)

conf.level = 0.95, ...)
modavgPred 157

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, type = "response", c.hat = 1, ...)

conf.level = 0.95, type = "response", c.hat = 1, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

modavgPred(cand.set, modnames = NULL,

conf.level = 0.95, ...)


158 modavgPred









modavgPred 159



Arguments
models.
newdata a data frame with the same structure as that of the original data frame for which
we want to make predictions.
rion (i.e., AICc).
value of nobs.
uncond.se either, old, or revised, specifying the equation used to compute the uncondi-
tional standard error of a model-averaged estimate. With uncond.se = "old",
was the former way to compute unconditional standard errors. With uncond.se = "revised",
equation 6.12 of Burnham and Anderson (2002) is used. Anderson (2008, p.
111) recommends use of the revised version for the computation of uncondi-
tional standard errors and it is now the default. Note that versions of package
AICcmodavg < 1.04 used the old method to compute unconditional standard
errors.
type the scale of prediction requested, one of response or link. The latter is only
relevant for glm, mer, and unmarkedFit classes. Note that the value terms is
160 modavgPred
not defined for modavgPred.

ure) syntax), with Poisson GLM’s, single-season and dynamic occupancy mod-
els (MacKenzie et al. 2002, 2003), or N-mixture models (Royle 2004, Dail and
Madsen 2011). If c.hat > 1, modavgPred will return the quasi-likelihood ana-
logue of the information criteria requested and multiply the variance-covariance
matrix of the estimates by this value (i.e., SE’s are multiplied by sqrt(c.hat)).
This option is not supported for generalized linear mixed models of the mer
class.
gamdisp the value of the gamma dispersion parameter.
unmarkedFitPCO, unmarkedFitDS, unmarkedFitGDS, unmarkedFitGMM, and unmarkedFitGPC
classes. The character strings supported vary with the type of model fitted.
For unmarkedFitOccu objects, either psi or detect can be supplied to indi-
cate whether the parameter is on occupancy or detectability, respectively. For
unmarkedFitColExt, possible values are psi, gamma, epsilon, and detect,
for parameters on occupancy in the inital year, colonization, extinction, and
detectability, respectively. For unmarkedFitOccuFP objects, one can specify
psi, detect, falsepos, and certain, for occupancy, detectability, probability
of assigning false-positives, and probability detections are certain, respectively.
enter lambda, gamma, omega, iota, or detect, to specify parameters on abun-
dance, recruitment, apparent survival, immigration, and detectability, respec-
tively. For unmarkedFitDS objects, lambda and detect are supported. For
unmarkedFitGDS, lambda, phi, and detect denote abundance, availability, and
detectability, respectively. For unmarkedFitGMM and unmarkedFitGPC objects,
lambda, phi, and detect denote abundance, availability, and detectability, re-
spectively.
Details
The candidate models must be stored in a list. Note that a data frame from which to make predictions
must be supplied with the newdata argument and that all variables appearing in the model set must
appear in this data frame. Variables must be of the same type as in the original analysis (e.g., factor,
numeric).
One can compute unconditional confidence intervals around the predictions from the elements re-
turned by modavgPred. The classic computation based on asymptotic normality of the estimator is
appropriate to estimate confidence intervals on the linear predictor (i.e., link scale). For predictions
of some types of response variables such as counts or binary variables, the normal approximation
modavgPred 161
may be inappropriate. In such cases, it is often better to compute the confidence intervals on the
linear predictor scale and then back-transform the limits to the scale of the response variable. These
are the confidence intervals returned by modavgPred. Burnham et al. (1987), Burnham and Ander-
son (2002, p. 164), and Williams et al. (2002) suggest alternative methods of computing confidence
intervals for small degrees of freedom with profile likelihood intervals or bootstrapping, but these
approaches are not yet implemented in modavgPred.
Value
modavgPred returns an object of class modavgPred with the following components:
type the scale of predicted values (response or link) for glm, mer, merMod, or unmarkedFit
classes.
mod.avg.pred the model-averaged prediction over the entire candidate model set.
uncond.se the unconditional standard error of each model-averaged prediction.
conf.level the confidence level used to compute the confidence interval.
lower.CL the lower confidence limit.
upper.CL the upper confidence limit.
matrix.output a matrix with rows consisting of the model-averaged predictions, the uncondi-
tional standard errors, and the confidence limits.
Author(s)
Marc J. Mazerolle
References
New York.
Burnham, K. P., Anderson, D. R., White, G. C., Brownie, C., Pollock, K. H. (1987) Design and
analysis methods for fish survival experiments based on release-recapture. American Fisheries
Society Monographs 5, 1–437.
2248–2255.
84, 2200–2207.
Williams, B. K., Nichols, J. D., Conroy, M. J. (2002) Analysis and Management of Animal Popula-
tions. Academic Press: New York.
162 modavgPred
See Also
AICc, aictab, importance, c_hat, confset, evidence, modavg, modavgCustom, modavgEffect,
modavgShrink, predict, predictSE
Examples
##example from subset of models in Table 1 in Mazerolle (2006)
data(dry.frog)

data = dry.frog)
##setup model names

##compute model-averaged value and unconditional SE of predicted log of

##mass lost for frogs of average mass in shade for each substrate type
##first create data set to use for predictions

new.dat <- data.frame(Shade = c(1, 1, 1),
cent_Initial_mass = c(0, 0, 0),
Initial_mass2 = c(0, 0, 0),
Substrate = c("SOIL", "SPHAGNUM", "PEAT"))
##compare unconditional SE's using both methods

modavgPred(cand.set = Cand.models, modnames = Modnames,
newdata = new.dat, type = "response", uncond.se = "old")
modavgPred(cand.set = Cand.models, modnames = Modnames,
newdata = new.dat, type = "response", uncond.se = "revised")
print(modavgPred(cand.set = Cand.models, modnames = Modnames,
newdata = new.dat, type = "response",
uncond.se = "revised"), digits = 4)
##Gamma glm
## Not run:
##clotting data example from 'gamma.shape' in MASS package of
modavgPred 163
##Venables and Ripley (2002, Modern applied statistics with

##S. Springer-Verlag: New York.)
clotting <- data.frame(u = c(5, 10, 15, 20, 30, 40, 60, 80, 100),
lot1 = c(118, 58, 42, 35, 27, 25, 21, 19, 18),
lot2 = c(69, 35, 26, 21, 18, 16, 13, 12, 12))
clot1 <- glm(lot1 ~ log(u), data = clotting, family = Gamma)
require(MASS)
gamma.dispersion(clot1) #dispersion parameter
gamma.shape(clot1) #reciprocal of dispersion parameter ==
##shape parameter
summary(clot1, dispersion = gamma.dispersion(clot1)) #better
##create list with models

Cand <- list( )
Cand[[1]] <- glm(lot1 ~ log(u), data = clotting, family = Gamma)
Cand[[2]] <- glm(lot1 ~ 1, data = clotting, family = Gamma)
##create vector of model names

Modnames <- paste("mod", 1:length(Cand), sep = "")
##compute model-averaged predictions on scale of response variable for

##all observations
modavgPred(cand.set = Cand, modnames = Modnames, newdata = clotting,
gamdisp = gamma.dispersion(clot1), type = "response")
##compute model-averaged predictions on scale of linear predictor

gamdisp = gamma.dispersion(clot1), type = "link")
##compute model-averaged predictions on scale of linear predictor

gamdisp = gamma.dispersion(clot1), type = "terms") #returns an error
##because type = "terms" is not defined for 'modavgPred'

type = "terms") #returns an error because
##no gamma dispersion parameter was specified (i.e., 'gamdisp' missing)
## End(Not run)
##example of model-averaged predictions from N-mixture model

##each variable appears twice in the models - this is a bit longer
## Not run:
require(unmarked)
data(mallard)
K = 30)
164 modavgPred
K = 30)
K = 30)
fm.mall.four <- pcount(~ ivel + date ~ 1, mallardUMF, K = 30)
##model list
Cands <- list(fm.mall.one, fm.mall.two, fm.mall.three, fm.mall.four)
Modnames <- c("length + forest", "elev + forest", "length + elev",
"null")
##compute model-averaged predictions of abundance for values of elev

modavgPred(cand.set = Cands, modnames = Modnames, newdata =
data.frame(elev = seq(from = -1.4, to = 2.4, by = 0.1),
length = 0, forest = 0), parm.type = "lambda",
type = "response")
##compute model-averaged predictions of detection for values of ivel

modavgPred(cand.set = Cands, modnames = Modnames, newdata =
data.frame(ivel = seq(from = -1.75, to = 5.9, by = 0.5),
date = 0), parm.type = "detect",
type = "response")
## End(Not run)
##example of model-averaged abundance from distance model

## Not run:
data(linetran) #example from ?distsamp

dist.breaks = c(0, 5, 10, 15, 20),
unitsIn = "m")
})

## Halfnormal. Covariates affecting both density and and detection.

fm2 <- distsamp(~area + habitat ~ habitat, ltUMF)
## Hazard function. Covariates affecting both density and and detection.

fm3 <- distsamp(~area + habitat ~ habitat, ltUMF, keyfun="hazard")

Modnames <- paste("mod", 1:length(Cands), sep = "")
##model-average predictions on abundance

modavgShrink 165
modavgPred(cand.set = Cands, modnames = Modnames, parm.type = "lambda", type = "link",

newdata = data.frame(area = mean(linetran$area), habitat = c("A", "B")))
## End(Not run)
##example using Orthodont data set from Pinheiro and Bates (2000)
## Not run:
require(nlme)

m2 <- gls(distance ~ 1, correlation = corCompSymm(value = 0.5, form = ~ 1 | Subject),

##assemble in list
Cand.models <- list(m1, m2)
##model names
Modnames <- c("age effect", "null model")

aictab(cand.set = Cand.models, modnames = Modnames)
##model-averaged predictions
modavgPred(cand.set = Cand.models, modnames = Modnames, newdata =
data.frame(age = c(8, 10, 12, 14)))
## End(Not run)
modavgShrink Compute Model-averaged Parameter Estimate with Shrinkage (Multi-

model Inference)
Description
This function computes an alternative version of model-averaging parameter estimates that consists
in shrinking estimates toward 0 to reduce model selection bias as in Burnham and Anderson (2002,
p. 152), Anderson (2008, pp. 130-132) and Lukacs et al. (2010). Specifically, models without the
parameter of interest have an estimate and variance of 0. modavgShrink also returns unconditional
standard errors and unconditional confidence intervals as described in Buckland et al. (1997) and
Burnham and Anderson (2002).
Usage
modavgShrink(cand.set, parm, modnames = NULL, second.ord = TRUE,
166 modavgShrink

...)
modavgShrink(cand.set, parm, modnames = NULL,
conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, c.hat = 1, gamdisp = NULL, ...)

conf.level = 0.95, ...)

modavgShrink 167

conf.level = 0.95, c.hat = 1, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, c.hat = 1, ...)
168 modavgShrink

modavgShrink(cand.set, parm, modnames =
conf.level = 0.95, c.hat = 1, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, ...)

conf.level = 0.95, c.hat = 1, ...)

conf.level = 0.95, ...)

conf.level = 0.95, c.hat = 1, parm.type = NULL, ...)

modavgShrink(cand.set, parm, modnames
= NULL, second.ord = TRUE, nobs = NULL, uncond.se =
"revised", conf.level = 0.95, c.hat = 1, parm.type = NULL,
...)

modavgShrink 169

= NULL, second.ord = TRUE, nobs = NULL, uncond.se = "revised",









170 modavgShrink
Arguments
parm the parameter of interest, enclosed between quotes, for which a model-averaged
estimate is required. For a categorical variable, the label of the estimate must be
included as it appears in the output (see ’Details’ below).
models.
rion (i.e., AICc).
value of nobs.
standard errors.
mixture models (Royle 2004, Dail and Madsen 2011). If c.hat > 1, modavgShrink
unmarkedFitPCO, unmarkedFitDS, unmarkedFitGDS, unmarkedFitGMM, unmarkedFitGPC,
modavgShrink 171

iota, or detect, to specify parameters on abundance, recruitment, apparent
survival, immigration, and detectability, respectively. For unmarkedFitDS ob-
jects, lambda and detect are supported. For unmarkedFitGDS, lambda, phi,
and detect denote abundance, availability, and detectability, respectively. For
Details
The parameter for which a model-averaged estimate is requested must be specified with the parm
argument and must be identical to its label in the model output (e.g., from summary). For factors,
one must specify the name of the variable and the level of interest. The shrinkage version of model
averaging is only appropriate for cases where each parameter is given an equal weighting in the
model (i.e., each parameter must appear the same number of times in the models) and has the same
interpretation across all models. As a result, models with interaction terms or polynomial terms are
not supported by modavgShrink.
modavgShrink is implemented for a list containing objects of aov, betareg, clm, clmm, clogit,
coxme, coxph, glm, glmmTMB, gls, hurdle, lm, lme, lmekin, maxlikeFit, mer, glmerMod, lmerMod,
lmerModLmerTest, multinom, polr, rlm, survreg, vglm, zeroinfl classes as well as various
models of unmarkedFit classes.
Value
modavgShrink creates an object of class modavgShrink with the following components:
Parameter the parameter for which a model-averaged estimate with shrinkage was ob-
tained.
Mod.avg.table the model selection table based on models including the parameter of interest.
Mod.avg.beta the model-averaged estimate based on all models.
Uncond.SE the unconditional standard error for the model-averaged estimate (as opposed to
the conditional SE based on a single model).
172 modavgShrink
Author(s)
Marc J. Mazerolle
References
New York.
Lukacs, P. M., Burnham, K. P., Anderson, D. R. (2010) Model selection bias and Freedman’s para-
dox. Annals of the Institute of Statistical Mathematics 62, 117–125.
2248–2255.
84, 2200–2207.
See Also
AICc, aictab, c_hat, importance, confset, evidence, modavg, modavgCustom, modavgPred
Examples
##cement example in Burnham and Anderson 2002
data(cement)
##setup same model set as in Table 3.2, p. 102
Cand.models[[1]] <- lm(y ~ x1 + x2, data = cement)
Cand.models[[2]] <- lm(y ~ x1 + x2 + x4, data = cement)
Cand.models[[7]] <- lm(y ~ x1 + x2 + x3 + x4, data = cement)
modavgShrink 173
Cand.models[[10]] <- lm(y ~ x4, data = cement)

##vector of model names

Modnames <- paste("mod", 1:15, sep="")
##AICc
aictab(cand.set = Cand.models, modnames = Modnames)
##compute model-averaged estimate with shrinkage - each parameter

##appears 8 times in the models
modavgShrink(cand.set = Cand.models, modnames = Modnames, parm = "x1")
##compare against classic model-averaging

modavg(cand.set = Cand.models, modnames = Modnames, parm = "x1")
##note that model-averaged estimate with shrinkage is closer to 0 than
##with the classic version
##remove a few models from the set and run again

Cand.unbalanced <- Cand.models[-c(3, 14, 15)]
##set up model names

Modnames <- paste("mod", 1:length(Cand.unbalanced), sep="")
##issues an error because some parameters appear more often than others
## Not run: modavgShrink(cand.set = Cand.unbalanced,
modnames = Modnames, parm = "x1")
## End(Not run)
##example on Orthodont data set in nlme

## Not run:
require(nlme)

##age and sex parameters appear in the same number of models
##same number of models with and without these parameters
##random is ~ age | Subject as it is a grouped data frame
method = "ML")
Cand.models[[4]] <- lme(distance ~ Sex, data = Orthodont, random = ~ 1,
method = "ML")

174 modavgShrink
##compute importance values for age

imp.age <- importance(cand.set = Cand.models, parm = "age",
modnames = Modnames, second.ord = TRUE,
nobs = NULL)
##compute shrinkage version of model averaging on age

mod.avg.age.shrink <- modavgShrink(cand.set = Cand.models,
parm = "age", modnames = Modnames,
second.ord = TRUE, nobs = NULL)
##compute classic version of model averaging on age

mod.avg.age.classic <- modavg(cand.set = Cand.models, parm = "age",
modnames = Modnames, second.ord = TRUE,
nobs = NULL)
##correspondence between shrinkage version and classic version of

##model averaging
mod.avg.age.shrink$Mod.avg.beta/imp.age$w.plus
mod.avg.age.classic$Mod.avg.beta
## End(Not run)
##example of N-mixture model modified from ?pcount

## Not run:
require(unmarked)
data(mallard)
K = 30)
K = 30)
K = 30)
##model list and names

Cands <- list(fm.mall.one, fm.mall.two, fm.mall.three)
Modnames <- c("length + forest", "elev + forest", "length + elev")
##compute model-averaged estimate with shrinkage for elev on abundance

modavgShrink(cand.set = Cands, modnames = Modnames, parm = "elev",
parm.type = "lambda")
## End(Not run)
multComp 175
multComp Create Model Selection Tables based on Multiple Comparisons
Description
This function is an alternative to traditional multiple comparison tests in designed experiments. It
creates a model selection table based on different grouping patterns of a factor and computes model-
averaged predictions for each of the factor levels. The current version works with objects of aov,
glm, gls, lm, lme, mer, merMod, lmerModLmerTest, negbin, and rlm, survreg classes.
Usage
multComp(mod, factor.id, letter.labels = TRUE, second.ord = TRUE,
nobs = NULL, sort = TRUE, newdata = NULL, uncond.se = "revised",
conf.level = 0.95, correction = "none", ...)

multComp(mod, factor.id, letter.labels = TRUE,
second.ord = TRUE, nobs = NULL, sort = TRUE, newdata = NULL,
uncond.se = "revised", conf.level = 0.95, correction = "none",
...)

...)

...)

type = "response", c.hat = 1, gamdisp = NULL, ...)

...)
## S3 method for class 'negbin'

176 multComp

type = "response", ...)

...)



uncond.se = "revised", conf.level = 0.95,
correction = "none", type = "response", ...)

uncond.se = "revised", conf.level = 0.95,
correction = "none", ...)
Arguments
mod a model of one of the above-mentioned classes that includes at least one factor
as an explanatory variable.
factor.id the factor of interest, on which the groupings (multiple comparisons) are based.
The user must supply the name of the categorical variable between quotes as it
appears in the model formula.
letter.labels logical. If TRUE, letters are used as labels to denote the grouping structure. If
FALSE, numbers are used as group labels.
rion (i.e., AICc), otherwise returns Akaike’s Information Criterion (AIC).
multComp 177
is relevant only for certain types of models such as mixed models where sam-
ple size is not straightforward. In such cases, one might use total number of
values.
we want to make predictions. This data frame should hold all variables constant
other than the factor.id variable. All levels of the factor.id variables should
be included in the newdata data frame to get model-averaged predictions for
each level. If NULL, model-averaged predictions are computed for each level of
the factor.id variable while the values of the other explanatory variables are
taken from the first row of the original data set.
was the former way to compute unconditional standard errors. With uncond.se = "revised",
equation 6.12 of Burnham and Anderson (2002) is used. Anderson (2008, p.
111) recommends use of the revised version for the computation of uncondi-
tional standard errors and it is now the default. Note that versions of package
AICcmodavg < 1.04 used the old method to compute unconditional standard er-
rors.
confidence intervals around predicted values for each level of factor.id.
correction the type of correction applied to obtain confidence intervals for simultaneous
inference (i.e., corrected for multiple comparisons). Current corrections include
"none" for uncorrected unconditional confidence intervals, "bonferroni" for
Bonferroni-adjusted confidence intervals (Dunn 1961), and "sidak" for Sidak-
adjusted confidence intervals (Sidak 1967).
type the scale of prediction requested, one of "response" or "link". The latter is
only relevant for glm and mer classes. Note that the value "terms" is not defined
for multComp.
obtained from c_hat. Note that values of c.hat different from 1 are only ap-
propriate for binomial GLM’s with trials > 1 (i.e., success/trial or cbind(success,
failure) syntax) or with Poisson GLM’s. If c.hat > 1, multComp will return the
quasi-likelihood analogue of the information criterion requested. This option is
not supported for generalized linear mixed models of the mer class.
gamdisp the value of the gamma dispersion parameter in a gamma GLM.
Details
A number of pairwise comparison tests are available for traditional experimental designs, some
controlling for the experiment-wise error and others for comparison-wise errors (Day and Quinn
1991). With the advent of information-theoretic approaches, there has been a need for methods
178 multComp
analogous to multiple comparison tests in a model selection framework. Dayton (1998) and Burn-
ham et al. (2011) suggested using different parameterizations or grouping patterns of a factor to
perform multiple comparisons with model selection. As such, it is possible to assess the support in
favor of certain grouping patterns based on a factor.
For example, a factor with three levels has four possible grouping patterns: abc (all groups are
different), abb (the first group differs from the other two), aab (the first two groups differ from the
third), and aaa (all groups are equal). multComp implements such an approach by pooling groups
of the factor variable in a model and updating the model, for each grouping pattern possible. The
models are ranked according to one of four information criteria (AIC, AICc, QAIC, and QAICc),
and the labels in the table correspond to the grouping pattern. Note that the factor levels are sorted
according to their means for the response variable before being assigned to a group. The function
also returns model-averaged predictions and unconditional standard errors for each level of the
factor.id variable based on the support in favor of each model (i.e., grouping pattern).
The number of grouping patterns increases substantially with the number of factor levels, as 2k−1 ,
where k is the number of factor levels. multComp supports factors with a maximum of 6 levels.
Also note that multComp does not handle models where the factor.id variable is involved in an
interaction. In such cases, one should create the interaction variable manually before fitting the
model (see Examples).
multComp currently implements three methods of computing confidence intervals. The default un-
conditional confidence intervals do not account for multiple comparisons (correction = "none").
With a large number m of potential pairwise comparisons among levels of factor.id, there is an in-
creased risk of type I error. For m pairwise comparisons and a given α level, correction = "bonferroni"
computes the unconditional confidence intervals based on αcorr = m α
(Dunn 1961). When correction = "sidak",
1
multComp reports Sidak-adjusted confidence intervals, i.e., αcorr = 1 − (1 − α) m .
Value
multComp creates a list of class multComp with the following components:
factor.id the factor for which grouping patterns are investigated.

models a list with the output of each model representing a different grouping pattern for
the factor of interest.
model.names a vector of model names denoting the grouping pattern for each level of the
factor.
model.table the model selection table for the models corresponding to each grouping pattern
for the factor of interest.
ordered.levels the levels of the factor ordered according to the mean of the response variable.
The grouping patterns (and model names) in the model selection table are based
on the same order.
model.avg.est a matrix with the model-averaged prediction, unconditional standard error, and
confidence intervals for each level of the factor.
conf.level the confidence level used for the confidence intervals.
correction the type of correction applied to the confidence intervals to account for potential
pairwise comparisons.
multComp 179
Author(s)
Marc J. Mazerolle
References
Day, R. W., Quinn, G. P. (1989) Comparisons of treatments after an analysis of variance in ecology.
Ecological Monographs 59, 433–463.
Dayton, C. M. (1998) Information criteria for the paired-comparisons problem. American Statisti-
cian, 52 144–151.
Dunn, O. J. (1961) Multiple comparisons among means. Journal of the American Statistical Asso-
ciation 56, 52–64.
Sidak, Z. (1967) Rectangular confidence regions for the means of multivariate normal distributions.
Journal of the American Statistical Association 62, 626–633.
See Also
aictab, confset, c_hat, evidence, glht, fit.contrast
Examples
##one-way ANOVA example
data(turkey)
##convert diet to factor

turkey$Diet <- as.factor(turkey$Diet)
##run one-way ANOVA
m.aov <- lm(Weight.gain ~ Diet, data = turkey)
##compute models with different grouping patterns

##and also compute model-averaged group means
out <- multComp(m.aov, factor.id = "Diet", correction = "none")
##look at results
out
##look at grouping structure of a given model

##and compare with original variable
cbind(model.frame(out$models[[2]]), turkey$Diet)
##evidence ratio
evidence(out$model.table)
##compute Bonferroni-adjusted confidence intervals

multComp(m.aov, factor.id = "Diet", correction = "bonferroni")

## Not run:
180 multComp
data(calcium)
m.aov2 <- lm(Calcium ~ Hormone + Sex + Hormone:Sex, data = calcium)
##multiple comparisons
multComp(m.aov2, factor.id = "Hormone")
##returns an error because 'Hormone' factor is
##involved in an interaction
##create interaction variable

calcium$inter <- interaction(calcium$Hormone, calcium$Sex)
##run model with interaction

m.aov.inter <- lm(Calcium ~ inter, data = calcium)
##compare both
logLik(m.aov2)
logLik(m.aov.inter)
##both are identical
##multiple comparisons
multComp(m.aov.inter, factor.id = "inter")
## End(Not run)
##Poisson regression
## Not run:
##example from ?glm
##Dobson (1990) Page 93: Randomized Controlled Trial :
counts <- c(18,17,15,20,10,20,25,13,12)
outcome <- gl(3,1,9)
treatment <- gl(3,3)
print(d.AD <- data.frame(treatment, outcome, counts))
glm.D93 <- glm(counts ~ outcome + treatment, data = d.AD, family = poisson)
multComp(mod = glm.D93, factor.id = "outcome")
## End(Not run)
##example specifying 'newdata'

## Not run:
data(dry.frog)
m1 <- lm(log_Mass_lost ~ Shade + Substrate +
data = dry.frog)
multComp(m1, factor.id = "Substrate",

newdata = data.frame(
Substrate = c("PEAT", "SOIL", "SPHAGNUM"),
Shade = 0, cent_Initial_mass = 0,
Initial_mass2 = 0))
newt 181
## End(Not run)
newt Newt Capture-mark-recapture Data
Description
This is a capture-mark-recapture data set on adult male and female Red-spotted Newts (Notoph-
thalmus viridescens) recorded by Gill (1985). A total of 1079 unique individuals were captured in
pitfall traps at a breeding site (White Oak Flat pond, Virginia) between 1975 and 1983.
Usage
data(newt)
Format
T1975 a binary variable, either 1 (captured) or 0 (not captured) during the 1975 breeding season.
Males a numeric variable indicating the total number of males with a given capture history.
Females a numeric variable indicating the total number of females with a given capture history.
Details
A single cohort of individuals was followed throughout the study, as all individuals were marked
in 1975 and no new individuals were added during the subsequent years. This data set is used to
illustrate classic Cormack-Jolly-Seber and related models (Cormack 1964, Jolly 1965, Seber 1965,
Lebreton et al. 1992, Mazerolle 2015).
182 Nmix.gof.test
Source
Cormack, R. M. (1964) Estimates of survival from the sighting of marked animals. Biometrika 51,
429–438.
Gill, D. E. (1985) Interpreting breeding patterns from census data: a solution to the Husting
dilemma. Ecology 66, 344–354.
Jolly, G. M. (1965) Explicit estimates from capture-recapture data with both death and immigration:
stochastic model. Biometrika 52, 225–247.
Laake, J. L. (2013) RMark: an R interface for analysis of capture-recapture data with MARK.
Alaska Fisheries Science Center (AFSC), National Oceanic and Atmospheric Administration, Na-
tional Marine Fisheries Service, AFSC Report 2013-01.
Monographs 62, 67-118.
Seber, G. A. F. (1965) A note on the multiple-recapture census. Biometrika 52, 249–259.
Examples
data(newt)
str(newt)
##convert raw capture data to capture histories

captures <- newt[, c("T1975", "T1976", "T1977", "T1978", "T1979",
"T1980", "T1981", "T1982", "T1983")]
newt.ch <- apply(captures, MARGIN = 1, FUN = function(i)
paste(i, collapse = ""))
##organize as a data frame readable by RMark package (Laake 2013)

##RMark requires at least one column called "ch"
##and another "freq" if summarized captures are provided
newt.full <- data.frame(ch = rep(newt.ch, 2),
freq = c(newt$Males, newt$Females),
Sex = c(rep("male", length(newt.ch)),
rep("female", length(newt.ch))))
str(newt.full)
newt.full$ch <- as.character(newt.full$ch)
##delete rows with 0 freqs

newt.full.orig <- newt.full[which(newt.full$freq != 0), ]
Nmix.gof.test Compute Chi-square Goodness-of-fit Test for N-mixture Models
Description
These functions compute a goodness-of-fit test for N-mixture models based on Pearson’s chi-square.
Nmix.gof.test 183
Usage
##methods for 'unmarkedFitPCount', 'unmarkedFitPCO',
##'unmarkedFitDS', 'unmarkedFitGDS', 'unmarkedFitGMM',
##'unmarkedFitGPC', and 'unmarkedFitMPois' classes
Nmix.chisq(mod, ...)
Nmix.gof.test(mod, nsim = 5, plot.hist = TRUE, report = NULL,

parallel = TRUE, ...)
Arguments
mod the N-mixture model of ’unmarkedFitPCount’, ’unmarkedFitPCO’, ’unmarked-
FitDS’, ’unmarkedFitGDS’, ’unmarkedFitGMM’, ’unmarkedFitGPC’, or ’un-
markedFitMPois’ classes for which a goodness-of-fit test is required.
nsim the number of bootstrapped samples.
plot.hist logical. Specifies that a histogram of the bootstrapped test statistic is to be in-
cluded in the output.
report If NULL, the test statistic for each iteration is not printed in the terminal. Other-
wise, an integer indicating the number of values of the test statistic that should
be printed on the same line. For example, if report = 3, the values of the test
statistic for three iterations are reported on each line.
parallel logical. If TRUE, requests that parboot use multiple cores to accelerate compu-
tations of the bootstrap.
Details
The Pearson chi-square can be used to assess the fit of N-mixture models. Instead of relying on the
theoretical distribution of the chi-square, a parametric bootstrap approach is implemented to obtain
P-values with the parboot function of the unmarked package. Nmix.chisq computes the observed
chi-square statistic based on the observed and expected counts from the model. Nmix.gof.test
calls internally Nmix.chisq and parboot to generate simulated data sets based on the model and
compute the chi-square test statistic.
It is also possible to obtain an estimate of the overdispersion parameter (c-hat) for the model at hand
by dividing the observed chi-square statistic by the mean of the statistics obtained from simulation
(MacKenzie and Bailey 2004, McKenny et al. 2006). This method of estimating c-hat is similar to
the one implemented for capture-mark-recapture models in program MARK (White and Burnham
1999).
Note that values of c-hat > 1 indicate overdispersion (variance > mean). Values much higher than
1 (i.e., > 4) probably indicate lack-of-fit. In cases of moderate overdispersion, one can multiply
the variance-covariance matrix of the estimates by c-hat. As a result, the SE’s of the estimates are
inflated (c-hat is also known as a variance inflation factor).
In model selection, c-hat should be estimated from the global model and the same value of c-hat
applied to the entire model set. Specifically, a global model is the most complex model which can
be simplified to yield all the other (nested) models of the set. When no single global model exists
184 Nmix.gof.test
in the set of models considered, such as when sample size does not allow a complex model, one can
estimate c-hat from ’subglobal’ models. Here, ’subglobal’ models denote models from which only
a subset of the models of the candidate set can be derived. In such cases, one can use the smallest
value of c-hat for model selection (Burnham and Anderson 2002).
in package AICcmodavg automatically add 1 when the c.hat argument > 1 and apply the same value
of c-hat for the entire model set. When c-hat > 1, functions compute quasi-likelihood information
criteria (either QAICc or QAIC, depending on the value of the second.ord argument) by scaling the
log-likelihood of the model by c-hat. The value of c-hat can influence the ranking of the models: as
c-hat increases, QAIC or QAICc will favor models with fewer parameters. As an additional check
against this potential problem, one can generate several model selection tables by incrementing
values of c-hat to assess the model selection uncertainty. If ranking changes only slightly up to the
c-hat value observed, one can be confident in making inference.
In cases of underdispersion (c-hat < 1), it is recommended to keep the value of c-hat to 1. However,
note that values of c-hat « 1 can also indicate lack-of-fit and that an alternative model should be
investigated.
Value
Nmix.chisq returns two value:

model.type the class of the fitted model.
Nmix.gof.test returns the following components:
model.type the class of the fitted model.

t.star the bootstrapped chi-square test statistics (i.e., obtained for each of the simulated
data sets).
p.value the P-value assessed from the parametric bootstrap, computed as the proportion
of the simulated test statistics greater than or equal to the observed test statistic.
c.hat.est the estimate of the overdispersion parameter, c-hat, computed as the observed
test statistic divided by the mean of the simulated test statistics.
nsim the number of bootstrap samples. The recommended number of samples varies
with the data set, but should be on the order of 1000 or 5000, and in cases with
a large number of visits, even 10 000 samples, namely to reduce the effect of
unusually small values of the test statistics.
Author(s)
Marc J. Mazerolle
References
pine 185
MacKenzie, D. I., Bailey, L. L. (2004) Assessing the fit of site-occupancy models. Journal of
Agricultural, Biological, and Environmental Statistics 9, 300–318.
McKenny, H. C., Keeton, W. S., Donovan, T. M. (2006). Effects of structural complexity enhance-
ment on eastern red-backed salamander (Plethodon cinereus) populations in northern hardwood
forests. Forest Ecology and Management 230, 186–196.
White, G. C., Burnham, K. P. (1999). Program MARK: Survival estimation from populations of
marked animals. Bird Study 46 (Supplement), 120–138.
See Also
AICc, c_hat, evidence, modavg, importance, mb.gof.test, modavgPred, pcount, pcountOpen,

parboot
Examples
##N-mixture model example modified from ?pcount
## Not run:
require(unmarked)
##single season
data(mallard)
##run model
fm.mallard <- pcount(~ ivel+ date + I(date^2) ~ length + elev + forest,
mallardUMF, K=30)
##compute observed chi-square

obs <- Nmix.chisq(fm.mallard)
obs

print(obs, digits.vals = 4)
##compute observed chi-square, assess significance, and estimate c-hat

obs.boot <- Nmix.gof.test(fm.mallard, nsim = 10)
##(e.g., 1000, 5000, or 10 000)
obs.boot
print(obs.boot, digits.vals = 4, digits.chisq = 4)
## End(Not run)
pine Strength of Pine Wood Based on the Density Adjusted for Resin Con-
tent
186 predictSE
Description
This data set consists of the strength of pine wood as a function of density or density adjusted for
resin content.
Usage
data(pine)
Format
y pine wood strength.

x pine wood density.
z pine wood density adjusted for resin content.
Details
Burnham and Anderson (2002, p. 183) use this data set originally from Carlin and Chib (1995) to
illustrate model selection for two competing and non-nested models.
Source
Carlin, B. P., Chib, S. (1995) Bayesian model choice via Markov chain Monte Carlo methods.
Journal of the Royal Statistical Society, Series B 57, 473–484.
Examples
data(pine)
## maybe str(pine) ; plot(pine) ...
predictSE Computing Predicted Values and Standard Errors
Description
Function to compute predicted values based on linear predictor and associated standard errors from
various fitted models.
predictSE 187
Usage
predictSE(mod, newdata, se.fit = TRUE, print.matrix = FALSE, ...)

predictSE(mod, newdata, se.fit = TRUE, print.matrix =
FALSE, ...)

FALSE, level = 0, ...)

FALSE, level = 0, type = "response", ...)

FALSE, level = 0, type = "response", ...)

FALSE, level = 0, ...)

predictSE(mod, newdata, se.fit = TRUE,
print.matrix = FALSE, type = "response", c.hat = 1, parm.type =
"lambda", ...)

predictSE(mod, newdata, se.fit = TRUE,
print.matrix = FALSE, type = "response", c.hat = 1,
parm.type = "lambda", ...)
Arguments
mod an object of class gls, lme, mer, merMod, lmerModLmerTest, unmarkedFitPCount,
or unmarkedFitPCO containing the output of a model.
we want to make predictions.
se.fit logical. If TRUE, compute standard errors on predictions.
print.matrix logical. If TRUE, the output is returned as a matrix, with predicted values and
standard errors in columns. If FALSE, the output is returned as a list.
level the level for which predicted values and standard errors are to be computed.
The current version of the function only supports predictions for the populations
excluding random effects (i.e., level = 0).
type specifies the type of prediction requested. This argument can take the value
response or link, for predictions on the scale of the response variable or on
188 predictSE
the scale of the linear predictor, respectively.

obtained from Nmix.gof.test. If c.hat > 1, predictSE will multiply
the variance-covariance matrix of the predictions by this value (i.e., SE’s are
multiplied by sqrt(c.hat)). High values of c.hat (e.g., c.hat > 4) may
indicate that model structure is inappropriate.
parm.type the parameter for which predictions are made based on the N-mixture model of
class unmarkedFitPCount or unmarkedFitPCO classes.
Details
predictSE computes predicted values and associated standard errors. Standard errors are approx-
imated using the delta method (Oehlert 1992). Predictions and standard errors for objects of gls
class and mixed models of lme, mer, merMod, lmerModLmerTest classes exclude the correlation or
variance structure of the model.
predictSE computes predicted values on abundance and standard errors based on the estimates
from an unmarkedFitPCount or unmarkedFitPCO object. Currently, only predictions on abundance
(i.e., parm.type = "lambda") with the zero-inflated Poisson distribution is supported. For other
parameters or distributions for models of unmarkedFit classes, use predict from the unmarked
package.
Value
predictSE returns requested values either as a matrix (print.matrix = TRUE) or list (print.matrix = FALSE)
with components:
fit the predicted values.

se.fit the standard errors of the predicted values (if se.fit = TRUE).
Note
For standard errors with better properties, especially for small samples, one can opt for simulations
(see Gelman and Hill 2007), or nonparametric bootstrap (Efron and Tibshirani 1998).
Author(s)
Marc J. Mazerolle
References
Efron, B., Tibshirani, R. J. (1998) An Introduction to the Bootstrap. Chapman & Hall/CRC: New
York.
Gelman, A., Hill, J. (2007) Data Analysis Using Regression and Multilevel/Hierarchical Models.
Cambridge University Press: New York.
Oehlert, G. W. (1992) A note on the delta method. American Statistician 46, 27–29.
predictSE 189
See Also
gls, lme, glmer, simulate.merMod, boot, parboot, nonparboot, pcount, pcountOpen, unmarkedFit-class
Examples
##Orthodont data from Pinheiro and Bates (2000) revisited
## Not run:
require(nlme)
data = Orthodont, method= "ML")
##compare against lme fit

logLik(m1)
logLik(lme(distance ~ age, random = ~1 | Subject, data = Orthodont,
method= "ML"))
##both are identical
##compute predictions and SE's for different ages

predictSE(m1, newdata = data.frame(age = c(8, 10, 12, 14)))
## End(Not run)
##example with mallard data set from unmarked package

## Not run:
require(unmarked)
data(mallard)
##run model with zero-inflated Poisson abundance
fm.mall.one <- pcount(~ ivel + date ~ length + forest, mallardUMF, K=30,
mixture = "ZIP")
##make prediction
predictSE(fm.mall.one, type = "response", parm.type = "lambda",
newdata = data.frame(length = 0, forest = 0, elev = 0))
##compare against predict
predict(fm.mall.one, type = "state", backTransform = TRUE,
##add offset in model to scale abundance per transect length

fm.mall.off <- pcount(~ ivel + date ~ forest + offset(length), mallardUMF, K=30,
mixture = "ZIP")
##make prediction
predictSE(fm.mall.off, type = "response", parm.type = "lambda",
##compare against predict
predict(fm.mall.off, type = "state", backTransform = TRUE,
190 salamander
## End(Not run)
salamander Salamander Capture-mark-recapture Data
Description
This is a capture-mark-recapture data set on male and female Spotted Salamanders (Ambystoma
maculatum) recorded by Husting (1965). A total of 1244 unique individuals were captured in pitfall
traps at a breeding site between 1959 and 1963.
Usage
data(salamander)
Format
Males a numeric variable indicating the total number of males with a given capture history. Nega-
tive values indicate losses on capture (animals not released on last capture).
Females a numeric variable indicating the total number of females with a given capture history.
Negative values indicate losses on capture (animals not released on last capture).
Details
This data set is used to illustrate classic Cormack-Jolly-Seber and related models (Cormack 1964,
Jolly 1965, Seber 1965, Lebreton et al. 1992).
Source
Cormack, R. M. (1964) Estimates of survival from the sighting of marked animals. Biometrika 51,
429–438.
Husting, E. L. (1965) Survival and breeding structure in a population of Ambystoma maculatum.
Copeia 1965, 352–362.
Jolly, G. M. (1965) Explicit estimates from capture-recapture data with both death and immigration:
stochastic model. Biometrika 52, 225–247.
Laake, J. L. (2013) RMark: an R interface for analysis of capture-recapture data with MARK.
Alaska Fisheries Science Center (AFSC), National Oceanic and Atmospheric Administration, Na-
tional Marine Fisheries Service, AFSC Report 2013-01.
summaryOD 191
Monographs 62, 67-118.
Seber, G. A. F. (1965) A note on the multiple-recapture census. Biometrika 52, 249–259.
Examples
data(salamander)
str(salamander)
##convert raw capture data to capture histories

captures <- salamander[, c("T1959", "T1960", "T1961", "T1962", "T1963")]
salam.ch <- apply(captures, MARGIN = 1, FUN = function(i)
paste(i, collapse = ""))
##organize as a data frame readable by RMark package (Laake 2013)

##RMark requires at least one column called "ch"
##and another "freq" if summarized captures are provided
salam.full <- data.frame(ch = rep(salam.ch, 2),
freq = c(salamander$Males, salamander$Females),
Sex = c(rep("male", length(salam.ch)),
rep("female", length(salam.ch))))
str(salam.full)
salam.full$ch <- as.character(salam.full$ch)
##delete rows with 0 freqs

salam.full.orig <- salam.full[which(salam.full$freq != 0), ]
summaryOD Display Model Summary Corrected for Overdispersion
Description
This function displays the estimates of a model with standard errors corrected for overdispersion
for a variety of model classes. The output includes either confidence intervals based on the normal
approximation or Wald hypothesis tests corrected for overdispersion.
Usage
summaryOD(mod, c.hat = 1, conf.level = 0.95,
out.type = "confint", ...)


192 summaryOD










## S3 method for class 'glmerMod'


summaryOD 193


Arguments
mod an object of class glm, glmmTMB, maxlikeFit, mer, merMod, multinom, vglm,
and various unmarkedFit classes containing the output of a model.
obtained from c_hat, mb.gof.test, or Nmix.gof.test.
conf.level the confidence level (1 − α) requested for the computation of confidence inter-
vals.
out.type the type of summary requested for each parameter estimate. If out.type = "confint",
computes confidence intervals corrected for overdispersion, whereas out.type = "nhst"
conducts null-hypothesis statistical testing corrected for overdispersion.
Details
Overdispersion occurs when the variance in the data exceeds that expected from a theoretical dis-
tribution such as the Poisson or binomial (McCullagh and Nelder 1989, Burnham and Anderson
2002). When the model is correct, small values of c-hat (1 < c-hat < 4) can reflect minor devia-
tions from model assumptions (Burnham and Anderson 2002). In such cases, it is possible to adjust
standard errors of parameter estimates by multiplying with sqrt(c.hat) (McCullagh and Nelder
1989). This is the correction applied by summaryOD.
Depending on the type of summary requested, i.e., out.type = "confint" or out.type = "nhst",
summaryOD will return either confidence intervals based on the normal approximation or Wald tests
for each parameter estimate (Agresti 1990).
For binomial distributions, note that values of c.hat > 1 are only appropriate with trials > 1 (i.e.,
success/trial or cbind(success, failure) syntax). The function supports different model
types such as Poisson GLM’s and GLMM’s, single-season occupancy models (MacKenzie et al.
2002), dynamic occupancy models (MacKenzie et al. 2003), or N-mixture models (Royle 2004,
Dail and Madsen 2011).
Value
summaryOD returns an object of class summaryOD as a list with the following components:
out.type the type of output requested by the user.
c.hat the c.hat estimate used to adjust standard errors.
conf.level the confidence level used to compute confidence intervals around the estimates.
outMat the output of the model corrected for overdispersion organized in a matrix.
194 summaryOD
Author(s)
Marc J. Mazerolle
References
Agresti, A. (2002) Categorical Data Analysis. Second edition. John Wiley and Sons: New Jersey.
2248–2255.
84, 2200–2207.
Hall: New York.
See Also
c_hat, mb.gof.test, Nmix.gof.test, anovaOD
Examples
data(min.trap)
##run model
m1 <- glm(Num_anura ~ Type + log.Perimeter + Num_ranatra,
data = min.trap)

c_hat(m1) #uses Pearson's chi-square/df
##display results corrected for overdispersion

summaryOD(m1, c_hat(m1))
summaryOD(m1, c_hat(m1), out.type = "nhst")
##example with occupancy model

## Not run:
tortoise 195

data(bullfrog)
##detection data
##run model
fm <- occu(~ 1 ~ 1, data = bfrog)
##check GOF
##GOF <- mb.gof.test(fm, nsim = 1000)
##estimate of c-hat: 1.89
##display results after overdispersion adjustment

summaryOD(fm, c.hat = 1.89)
summaryOD(fm, c.hat = 1.89, out.type = "nhst")
}
## End(Not run)
tortoise Gopher Tortoise Distance Sampling Data
Description
This simulated data set by Mazerolle (2015) is based on the biological parameters for the Gopher
Tortoise (Gopherus polyphemus) reported by Smith et al. (2009). A half-normal distribution with a
scale of 10 and without an adjustment factor was used to simulate the distance data for a study area
of 120 km2 . An effort of 500 m in 300 line transects was deployed. A density of 72 individuals per
km2 was used in the simulation using the approach outlined in Buckland et al. (2001).
Usage
data(tortoise)
Format
Region.Label a numeric identifier for the study area.

Area a numeric variable for the surface area of the study area in square meters.
196 turkey
Sample.Label a numeric identifier for each line transect relating each observation to its corre-
sponding transect.
Effort Effort in meters expended in each line transect.
distance a numeric variable for the perpendicular distances in meters relative to the transect line
for each of the individuals detected during the survey. Note that transects without detections
have a value of NA for this variable.
Details
This data set is used to illustrate classic distance sampling (Buckland et al. 2001, Mazerolle 2015).
Source
Buckland, S. T., Anderson, D. R., Burnham, K. P., Laake, J. L., Borchers, D. L., Thomas, L.
(2001) Introduction to distance sampling: estimating abundance of biological populations. Oxford
University Press: Oxford.
Smith, L. L., Linehan, J. M., Stober, J. M., Elliott, M. J., Jensen, J. B. (2009) An evaluation of
distance sampling for large-scale gopher tortoise surveys in Georgia, USA. Applied Herpetology 6,
355–368.
Examples
data(tortoise)
str(tortoise)
##plot distance data to determine if truncation is required

##(Buckland et al. 2001, pp. 15--17)
hist(tortoise$distance)
turkey Turkey Weight Gain
Description
This one-way ANOVA data set presents turkey weight gain in pounds across five diets.
Usage
data(turkey)
Format
Diet diet factor with 5 levels.
Weight.gain weight gain in pounds.
useBIC 197
Details
Heiberger and Holland (2004) and Ott (1993) analyze this data set to illustrate one-way ANOVA.
Source
Heiberger, R. M., Holland, B. (2004) Statistical Analysis and Data Display: an intermediate course
with examples in S-Plus, R, and SAS. Springer: New York.
Ott, R. L. (1993) An Introduction to Statistical Methods and Data Analysis. Fourth edition. Duxbury:
Pacific Grove, CA.
Examples
data(turkey)
str(turkey)
useBIC Computing BIC or QBIC
Description
Functions to compute the Bayesian information criterion (BIC) or a quasi-likelihood analogue
(QBIC).
Usage
useBIC(mod, return.K = FALSE, nobs = NULL, ...)






## S3 method for class 'fitdist'

198 useBIC
## S3 method for class 'fitdistr'


useBIC(mod, return.K = FALSE, nobs = NULL, c.hat = 1,
...)

...)








useBIC(mod, return.K = FALSE, nobs = NULL, c.hat =
1, ...)




useBIC 199
...)






useBIC(mod, return.K = FALSE, nobs = NULL, c.hat =
1, ...)

...)

Arguments
mod an object of class aov, betareg, clm, clmm, clogit, coxme, coxph, fitdist,
fitdistr, glm, glmmTMB, gls, gnls, hurdle, lavaan, lm, lme, lmekin, maxlikeFit,
mer, merMod, lmerModLmerTest, multinom, nlme, nls, polr, rlm, survreg,
vglm, zeroinfl, and various unmarkedFit classes containing the output of a
model.
compute the BIC (i.e., nobs defaults to total number of observations). This is
relevant only for mixed models or various models of unmarkedFit classes where
sample size is not straightforward. In such cases, one might use total number of
mixture models (Royle 2004, Dail and Madsen 2011). If c.hat > 1, useBIC
200 useBIC

Details
useBIC computes the Bayesian information criterion (BIC, Schwarz 1978):
BIC = −2 ∗ log − likelihood + K ∗ log(n),
where the log-likelihood is the maximum log-likelihood of the model, K corresponds to the number
of estimated parameters, and n corresponds to the sample size of the data set.
In the presence of overdispersion, a quasi-likelihood analogue of the BIC (QBIC) will be computed,
as
QBIC = + K ∗ log(n),
c − hat
where c-hat is the overdispersion parameter specified by the user with the argument c.hat. Note
that BIC or QBIC values are meaningful to select among gls or lme models fit by maximum like-
lihood. BIC or QBIC based on REML are valid to select among different models that only differ in
their random effects (Pinheiro and Bates 2000).
Value
useBIC returns the BIC or the number of estimated parameters, depending on the values of the
arguments.
Note
The actual (Q)BIC values are not really interesting in themselves, as they depend directly on the
Author(s)
Marc J. Mazerolle
References
2248–2255.
useBICCustom 201
84, 2200–2207.
Pinheiro, J. C., Bates, D. M. (2000) Mixed-effect models in S and S-PLUS. Springer Verlag: New
York.
See Also
AICc, bictab, bictabCustom, useBICCustom
Examples
data(cement)
##compute BIC with full likelihood

useBIC(glob.mod, return.K = FALSE)
##compute BIC for mixed model on Orthodont data set in Pinheiro and
##Bates (2000)
## Not run:
require(nlme)
m1 <- lme(distance ~ age, random = ~1 | Subject, data = Orthodont,
method= "ML")
useBIC(m1, return.K = FALSE)
## End(Not run)
useBICCustom Custom Computation of BIC and QBIC from User-supplied Input
Description
This function computes the Bayesian information criterion (BIC) or a quasi-likelihood counterpart
(QBIC) from user-supplied input instead of extracting the values automatically from a model object.
This function is particularly useful for output imported from other software or for model classes that
are not currently supported by useBIC.
Usage
useBICCustom(logL, K, return.K = FALSE, nobs = NULL, c.hat = 1)
202 useBICCustom
Arguments
logL the value of the model log-likelihood.
K the number of estimated parameters in the model.
nobs the sample size required to compute the BIC or QBIC.
c.hat > 1, useBICCustom will return the quasi-likelihood analogue of the infor-
Details
useBICCustom computes one of the following two information criteria:
the Bayesian information criterion (BIC, Schwarz 1978) or the quasi-likelihood BIC (QBIC).
Value
useBICCustom returns the BIC or QBIC depending on the values of the c.hat argument.
Note
The actual (Q)BIC values are not really interesting in themselves, as they depend directly on the
Author(s)
Marc J. Mazerolle
References
2248–2255.
xtable 203
84, 2200–2207.
See Also
AICc, aictabCustom, useBIC, bictab, evidence, modavgCustom
Examples
data(cement)
LL <- logLik(glob.mod)[1]
##extract number of parameters

##including residual variance
K.mod <- length(coef(glob.mod)) + 1
##compute BIC with full likelihood

useBICCustom(LL, K.mod, nobs = nrow(cement))
##compare against useBIC
useBIC(glob.mod)
xtable Format Objects to LaTeX or HTML
Description
Functions to format various objects following model selection and multimodel inference to LaTeX
or HTML tables. These functions extend the methods from the xtable package (Dahl 2014).
Usage
## S3 method for class 'aictab'
xtable(x, caption = NULL, label = NULL, align = NULL,
digits = NULL, display = NULL, nice.names = TRUE,
include.AICc = TRUE, include.LL = TRUE, include.Cum.Wt = FALSE,
...)
## S3 method for class 'anovaOD'

204 xtable
digits = NULL, display = NULL,

nice.names = TRUE, ...)
## S3 method for class 'bictab'

include.BIC = TRUE, include.LL = TRUE, include.Cum.Wt = FALSE,
...)
## S3 method for class 'boot.wt'

include.AICc = TRUE, include.AICcWt = FALSE, ...)
## S3 method for class 'countDist'

xtable(x, caption = NULL, label = NULL,
align = NULL, digits = NULL, display = NULL,
nice.names = TRUE, table.countDist = "distance", ...)
## S3 method for class 'checkParms'

nice.names = TRUE, include.variable = TRUE, include.max.se =
TRUE, include.n.high.se = TRUE, ...)
## S3 method for class 'countHist'

nice.names = TRUE, table.countHist = "count", ...)
## S3 method for class 'detHist'

nice.names = TRUE, table.detHist = "freq", ...)
## S3 method for class 'dictab'

include.DIC = TRUE, include.Cum.Wt = FALSE, ...)
## S3 method for class 'ictab'

include.IC = TRUE, include.Cum.Wt = FALSE, ...)
## S3 method for class 'mb.chisq'

xtable 205
include.detection.histories = TRUE, ...)
## S3 method for class 'modavg'

print.table = FALSE, ...)
## S3 method for class 'modavgCustom'

align = NULL, digits = NULL, display = NULL, nice.names = TRUE,
## S3 method for class 'modavgEffect'

## S3 method for class 'modavgIC'

## S3 method for class 'modavgPred'

...)
## S3 method for class 'modavgShrink'

## S3 method for class 'multComp'

## S3 method for class 'summaryOD'

digits = NULL, display = NULL,
nice.names = TRUE, ...)
Arguments
x an object of class aictab, anovaOD, bictab, boot.wt, checkParms, countDist,

countHist, detHist, dictab, ictab, mb.chisq, modavg, modavgEffect, modavgCustom,
modavgIC, modavgPred, modavgShrink, multComp, or summaryOD.
206 xtable
caption a character vector of length 1 or 2 storing the caption or title of the table. If
the vector is of length 2, the second item is the short caption used when LaTeX
generates a list of tables. The default value is NULL and suppresses the caption.
label a character vector storing the LaTeX label or HTML anchor. The default value
is NULL and suppresses the label.
align a character vector of length equal to the number of columns of the table speci-
fying the alignment of the elements. Note that the rownames are considered as
an additional column and require an alignment value.
digits a numeric vector of length one or equal to the number of columns in the table
(including the rownames) specifying the number of digits to display in each
column.
display a character vector of length equal to the number of columns (including the row-
names) specifying the format of each column. For example, use s for strings, f
for numbers in the regular format, or d for integers. See formatC for additional
possible values.
nice.names logical. If TRUE, column labels are modified to improve their appearance in the
table. If FALSE, simpler labels are used, or the ones supplied directly by the user
in the object storing the output.
include.AICc logical. If TRUE, the column containing the information criterion (AIC, AICc,
QAIC, or QAICc) of each model is printed in the table. If FALSE, the column is
suppressed.
include.BIC logical. If TRUE, the column containing the information criterion (BIC or QBIC)
of each model is printed in the table. If FALSE, the column is suppressed.
include.DIC logical. If TRUE, the column containing the deviance information criterion (DIC)
of each model is printed in the table. If FALSE, the column is suppressed.
include.IC logical. If TRUE, the column containing the information criterion of each model
is printed in the table. If FALSE, the column is suppressed.
include.LL logical. If TRUE, the column containing the log-likelihood of each model is
printed in the table. If FALSE, the column is suppressed.
include.Cum.Wt logical. If TRUE, the column containing the cumulative Akaike weights is printed
in the table. If FALSE, the column is suppressed.
include.AICcWt logical. If TRUE, the column containing the Akaike weight of each model is
printed in the table. If FALSE, the column is suppressed.
include.detection.histories
logical. If TRUE, the column containing detection histories is printed in the table.
If FALSE, the column is suppressed.
include.variable
logical. If TRUE, the column containing the variable name is printed in the table.
If FALSE, the column is suppressed.
include.max.se logical. If TRUE, the column containing the maximum SE in the model is printed
in the table. If FALSE, the column is suppressed.
include.n.high.se
logical. If TRUE, the column containing the number of SE’s greater than the
threshold specified by the user is printed in the table. If FALSE, the column is
suppressed.
xtable 207
print.table logical. If TRUE, the model selection table is printed and other sections of the
output are suppressed (e.g., model-averaged estimates). If FALSE, the model
selection table is suppressed and only the other portion of the output is printed
in the table.
table.detHist character string specifying, either "freq", "prop", or "hist". If table.type = "freq",
the function returns a table of frequencies of sites sampled, of sites with at least
one detection, and for data with multiple primary periods, the frequencies of
sites with observed extinctions and colonizations. If table.type = "prop",
the table returns the proportion of sites with at least one detection, and for data
with multiple periods, the proportion of sites with observed extinctions and col-
onizations. If table.type = "hist", the function returns the frequencies of
each observed detection history.
table.countDist
character string specifying, either "distance", "count", "freq", or "prop". If
table.type = "distance", the function returns a table of counts summarized
for each distance class. If table.type = "count", the function returns the
table of frequencies of counts observed across sites. If table.type = "freq",
the function returns a table of frequencies of sites sampled, of sites with at least
one detection, and for data with multiple primary periods, the frequencies of
sites with observed extinctions and colonizations. If table.type = "prop",
the table returns the proportion of sites with at least one detection, and for data
with multiple periods, the proportion of sites with observed extinctions and col-
onizations.
table.countHist
character string specifying, either "count", "freq", "prop", or "hist". If
table.type = "count", the function returns the table of frequencies of counts
observed across sites. If table.type = "freq", the function returns a
table of frequencies of sites sampled, of sites with at least one detection, and
for data with multiple primary periods, the frequencies of sites with observed
extinctions and colonizations. If table.type = "prop", the table returns the
proportion of sites with at least one detection, and for data with multiple pe-
riods, the proportion of sites with observed extinctions and colonizations. If
table.type = "hist", the function returns the frequencies of each observed
count history.
Details
xtable creates an object of the xtable class inheriting from the data.frame class. This object can
then be used with print.xtable for added flexibility such as suppressing row names, modifying
caption placement, and format tables in LaTeX or HTML format.
Author(s)
Marc J. Mazerolle
208 xtable
References
Dahl, D. B. (2014) xtable: Export tables to LaTeX or HTML. R package version 1.7-3. https:
//cran.r-project.org/package=xtable.
See Also
aictab, boot.wt, dictab, formatC, ictab, mb.chisq, modavg, modavgCustom, modavgIC, modavgEffect,
modavgPred, modavgShrink, multComp, summaryOD, anovaOD, xtable, print.xtable
Examples
##model selection example
data(dry.frog)
##setup candidate models
data = dry.frog)
Model.names <- c("additive", "interaction", "no shade")
##model selection table - AICc

out <- aictab(cand.set = Cand.models, modnames = Model.names)
xtable(out)
##exclude AICc and LL
xtable(out, include.AICc = FALSE, include.LL = FALSE)
##remove row names and add caption
print(xtable(out, caption = "Model selection based on AICc"),
include.rownames = FALSE, caption.placement = "top")
##model selection table - BIC

out2 <- bictab(cand.set = Cand.models, modnames = Model.names)
xtable(out2)
##exclude AICc and LL
xtable(out2, include.BIC = FALSE, include.LL = FALSE)
##remove row names and add caption
print(xtable(out2, caption = "Model selection based on BIC"),
include.rownames = FALSE, caption.placement = "top")
##model-averaged estimate of Initial_mass2

mavg.mass <- modavg(cand.set = Cand.models, parm = "Initial_mass2",
modnames = Model.names)
#model-averaged estimate
xtable 209
xtable(mavg.mass, print.table = FALSE)

#table with contribution of each model
xtable(mavg.mass, print.table = TRUE)
##model-averaged predictions for first 10 observations

preds <- modavgPred(cand.set = Cand.models, modnames = Model.names,
newdata = dry.frog[1:10, ])
xtable(preds)
}
##example of diagnostics
## Not run:
##distance sampling example from ?distsamp
data(linetran)
dist.breaks = c(0, 5, 10, 15, 20),
unitsIn = "m")
})
##summarize counts across distance classes

xtable(countDist(ltUMF), table.countDist = "distance")
##summarize counts across all sites
xtable(countDist(ltUMF), table.countDist = "count")
##Half-normal detection function

##determine parameters with highest SE's
xtable(checkParms(fm1))
}
## End(Not run)
Index
∗Topic datasets extractLL, 93

beetle, 39 extractSE, 94
bullfrog, 55 extractX, 96
calcium, 56 fam.link.mer, 102
cement, 57 ictab, 106
dry.frog, 87 importance, 108
fat, 103 mb.gof.test, 119
gpa, 105 modavg, 125
iron, 115 modavg.utility, 136
lizards, 116 modavgCustom, 139
min.trap, 124 modavgEffect, 142
newt, 181 modavgIC, 154
pine, 185 modavgPred, 156
salamander, 190 modavgShrink, 165
tortoise, 195 multComp, 175
turkey, 196 Nmix.gof.test, 182
∗Topic models predictSE, 186
AICc, 11 summaryOD, 191
AICcCustom, 16 useBIC, 197
useBICCustom, 201
AICcmodavg-defunct, 18
xtable, 203
AICcmodavg-package, 3
aictab, 23 AICc, 4, 11, 18, 30, 54, 67, 78, 89, 94, 113,
aictabCustom, 32 123, 133, 148, 162, 172, 185, 201,
anovaOD, 34 203
bictab, 40 AICc.mult (AICcmodavg-defunct), 18
bictabCustom, 48 AICc.unmarked (AICcmodavg-defunct), 18
boot.wt, 51 AICcCustom, 16, 16, 34, 50, 84, 141
c_hat, 76 AICcmodavg (AICcmodavg-package), 3
checkConv, 58 AICcmodavg-defunct, 18
checkParms, 60 AICcmodavg-package, 3
confset, 64 aictab, 4, 16, 22, 23, 46, 67, 84, 86, 89, 94,
countDist, 68 113, 133, 138, 148, 162, 172, 179,
countHist, 71 208
covDiag, 74 aictab.clm (AICcmodavg-defunct), 18
detHist, 79 aictab.clmm (AICcmodavg-defunct), 18
DIC, 83 aictab.coxph (AICcmodavg-defunct), 18
dictab, 85 aictab.glm (AICcmodavg-defunct), 18
evidence, 88 aictab.gls (AICcmodavg-defunct), 18
extractCN, 91 aictab.lm (AICcmodavg-defunct), 18
210
INDEX 211
aictab.lme (AICcmodavg-defunct), 18 extractCN, 5, 63, 91, 99

aictab.mer (AICcmodavg-defunct), 18 extractLL, 5, 22, 93
aictab.merMod (AICcmodavg-defunct), 18 extractSE, 5, 94, 99
aictab.mult (AICcmodavg-defunct), 18 extractX, 5, 96
aictab.nlme (AICcmodavg-defunct), 18
aictab.nls (AICcmodavg-defunct), 18 fam.link.mer, 5, 102
aictab.polr (AICcmodavg-defunct), 18 fat, 103
aictab.rlm (AICcmodavg-defunct), 18 fit.contrast, 179
aictab.unmarked (AICcmodavg-defunct), 18 formatC, 208
aictabCustom, 18, 30, 32, 50, 86, 107, 141, formatCands (modavg.utility), 136
155, 203 formulaShort (modavg.utility), 136
anovaOD, 5, 34, 78, 138, 194, 208
gdistsamp, 94
beetle, 39 glht, 179
bictab, 4, 30, 40, 89, 201, 203 glmer, 95, 102, 189
bictabCustom, 34, 46, 48, 141, 201 gls, 189
boot, 189 gpa, 105
boot.wt, 4, 51, 208
ictab, 34, 50, 106, 155, 208
bullfrog, 55
importance, 4, 16, 22, 30, 46, 54, 67, 78, 89,
108, 123, 133, 148, 162, 172, 185
c_hat, 5, 16, 18, 22, 30, 34, 38, 50, 54, 63, 67,
iron, 115
76, 89, 92, 113, 123, 133, 148, 162,
172, 179, 185, 194 kappa, 92
calcium, 56
cement, 57 lizards, 116
checkConv, 5, 58, 63 lme, 189
checkParms, 5, 59, 60 lmekin, 94, 95
colext, 94, 123 lmer, 95, 102
confset, 4, 16, 18, 22, 30, 34, 46, 50, 54, 64,
78, 84, 86, 89, 107, 113, 133, 148, maxlike, 94
162, 172, 179 mb.chisq, 82, 208
countDist, 5, 63, 68, 73, 82 mb.chisq (mb.gof.test), 119
countHist, 5, 63, 70, 71, 82 mb.gof.test, 5, 38, 59, 63, 78, 82, 92, 119,
covDiag, 5, 59, 70, 73, 74, 82 185, 194
coxme, 94, 95 min.trap, 124
coxph, 94 modavg, 4, 16, 22, 30, 54, 67, 75, 78, 89, 95,
102, 113, 123, 125, 138, 141, 155,
detHist, 5, 63, 70, 73, 79 162, 172, 185, 208
DIC, 4, 83, 86 modavg.clm (AICcmodavg-defunct), 18
dictab, 4, 22, 84, 85, 208 modavg.clmm (AICcmodavg-defunct), 18
dictab.bugs (AICcmodavg-defunct), 18 modavg.coxph (AICcmodavg-defunct), 18
dictab.rjags (AICcmodavg-defunct), 18 modavg.effect (AICcmodavg-defunct), 18
distsamp, 94 modavg.glm (AICcmodavg-defunct), 18
dry.frog, 87 modavg.gls (AICcmodavg-defunct), 18
modavg.lme (AICcmodavg-defunct), 18
evidence, 4, 16, 18, 22, 30, 34, 46, 50, 54, 67, modavg.mer (AICcmodavg-defunct), 18
78, 84, 86, 88, 107, 113, 123, 133, modavg.merMod (AICcmodavg-defunct), 18
148, 162, 172, 179, 185, 203 modavg.mult (AICcmodavg-defunct), 18
extract.LL (AICcmodavg-defunct), 18 modavg.polr (AICcmodavg-defunct), 18
212 INDEX
modavg.rlm (AICcmodavg-defunct), 18 print.evidence (evidence), 88

modavg.shrink (AICcmodavg-defunct), 18 print.extractCN (extractCN), 91
modavg.unmarked (AICcmodavg-defunct), 18 print.extractX (extractX), 96
modavg.utility, 136 print.ictab (ictab), 106
modavgCustom, 18, 34, 50, 99, 107, 133, 139, print.importance (importance), 108
155, 162, 172, 203, 208 print.mb.chisq (mb.gof.test), 119
modavgEffect, 4, 22, 30, 99, 133, 142, 162, print.modavg (modavg), 125
208 print.modavgCustom (modavgCustom), 139
modavgIC, 107, 141, 154, 208 print.modavgEffect (modavgEffect), 142
modavgPred, 4, 16, 22, 30, 54, 67, 75, 78, 89, print.modavgIC (modavgIC), 154
99, 102, 113, 123, 133, 138, 141, print.modavgPred (modavgPred), 156
148, 155, 156, 172, 185, 208 print.modavgShrink (modavgShrink), 165
modavgpred (AICcmodavg-defunct), 18 print.multComp (multComp), 175
modavgShrink, 4, 16, 22, 30, 54, 67, 89, 113, print.Nmix.chisq (Nmix.gof.test), 182
133, 138, 141, 148, 155, 162, 165, print.summaryOD (summaryOD), 191
208 print.xtable, 208
mult.comp (AICcmodavg-defunct), 18
multComp, 4, 22, 175, 208 rcond, 92
reverse.exclude (modavg.utility), 136
newt, 181 reverse.parm (modavg.utility), 136
Nmix.chisq, 70, 73, 75
Nmix.chisq (Nmix.gof.test), 182 salamander, 190
Nmix.gof.test, 5, 38, 59, 63, 70, 73, 75, 78, simulate.merMod, 189
92, 123, 182, 194 summaryOD, 5, 38, 78, 191, 208
nonparboot, 189
tortoise, 195
occu, 94, 123 turkey, 196
occuRN, 94
useBIC, 4, 16, 46, 197, 203
useBICCustom, 201, 201
parboot, 63, 92, 123, 185, 189
pcount, 75, 94, 185, 189 xtable, 5, 203, 208
pcountOpen, 94, 185, 189
pine, 185
predict, 99, 162
predictSE, 5, 22, 75, 99, 102, 162, 186
predictSE.zip (AICcmodavg-defunct), 18
print.aictab (aictab), 23
print.anovaOD (anovaOD), 34
print.bictab (bictab), 40
print.boot.wt (boot.wt), 51
print.c_hat (c_hat), 76
print.checkConv (checkConv), 58
print.checkParms (checkParms), 60
print.confset (confset), 64
print.countDist (countDist), 68
print.countHist (countHist), 71
print.covDiag (covDiag), 74
print.detHist (detHist), 79
print.dictab (dictab), 85

AICcmodavg PDF

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

AICcmodavg PDF

Caricato da

Copyright:

Formati disponibili

Package ‘AICcmodavg’

May 29, 2019

AICcmodavg-package Model Selection and Multimodel Inference Based on (Q)AIC(c)

• modavgCustom Computes model-averaged parameter estimate based on (Q)AIC(c) from user-

A number of functions for model diagnostics are available:

Other utility functions include:

##set up candidate models in a list

##check c-hat for global model

##output of model corrected for overdispersion

##assign names to each model

Modnames <- c("type + logperim + invertpred", "type + logperim",

##model selection table based on AICc

##compute evidence ratio

##compute confidence set based on 'raw' method

##compute importance value for "TypeBOG" - same number of models

##compute model-averaged estimate of "TypeBOG" using the natural average

##compute model-averaged estimate of "TypeBOG" using shrinkage estimator

##compute model-averaged predictions for two types of ponds

##model-averaged predictions across entire model set

##compute model-averaged effect size between two groups

## observation covariates are in site-major, observation-minor order

##check detection history data from data object

##set up candidate model set

fm2 <- occu(~ 1 ~ sitevar1, pferUMF)

##assign names to elements in list

##check GOF of global model and estimate c-hat

##check for high SE's in models

##export as LaTeX table

##compute evidence ratio

##compute confidence set based on 'raw' method

##compute importance value for "sitevar1" on occupancy

##compute model-averaged estimate of "sitevar1" on occupancy

modavg(cand.set = Cand.models, parm = "sitevar1",

##compute model-averaged estimate of "sitevar1"

##compute model-average predictions

##create a data set for predictions

##model-averaged predictions of psi across range of values

##vector with model LL's

##vector with number of parameters

##create a vector of names to trace back models in set

##generate AICc table

##model averaging parameter estimate

##vector of SE's of beta estimates for a parameter of interest

##compute model-averaged estimate and unconditional SE based on AICc

##create a vector of names to trace back models in set

##generate WAIC model selection table

##compute model-averaged estimate

##compute model-averaged estimate and unconditional SE based on WAIC

##export as LaTeX table

AICc Computing AIC, AICc, QAIC, and QAICc

## S3 method for class 'aov'

## S3 method for class 'betareg'

## S3 method for class 'clm'

## S3 method for class 'clmm'

## S3 method for class 'coxme'

## S3 method for class 'coxph'

## S3 method for class 'fitdist'

## S3 method for class 'fitdistr'

## S3 method for class 'glm'

## S3 method for class 'glmmTMB'

## S3 method for class 'gls'

## S3 method for class 'gnls'

## S3 method for class 'hurdle'